minix/drivers/at_wini/at_wini.c

2245 lines
62 KiB
C
Raw Normal View History

2005-04-21 16:53:53 +02:00
/* This file contains the device dependent part of a driver for the IBM-AT
* winchester controller. Written by Adri Koppes.
*
* Changes:
* Oct 2, 2013 drop non-PCI support; one controller per instance (David)
2005-10-20 22:26:02 +02:00
* Aug 19, 2005 ATA PCI support, supports SATA (Ben Gras)
2005-04-21 16:53:53 +02:00
* Nov 18, 2004 moved AT disk driver to user-space (Jorrit N. Herder)
* Aug 20, 2004 watchdogs replaced by sync alarms (Jorrit N. Herder)
* Mar 23, 2000 added ATAPI CDROM support (Michael Temari)
* May 14, 2000 d-d/i rewrite (Kees J. Bot)
* Apr 13, 1992 device dependent/independent split (Kees J. Bot)
*/
#include "at_wini.h"
#include <minix/sysutil.h>
#include <minix/type.h>
#include <minix/endpoint.h>
#include <sys/ioc_disk.h>
#include <machine/pci.h>
#include <sys/mman.h>
Driver refactory for live update and crash recovery. SYSLIB CHANGES: - DS calls to publish / retrieve labels consider endpoints instead of u32_t. VFS CHANGES: - mapdriver() only adds an entry in the dmap table in VFS. - dev_up() is only executed upon reception of a driver up event. INET CHANGES: - INET no longer searches for existing drivers instances at startup. - A newtwork driver is (re)initialized upon reception of a driver up event. - Networking startup is now race-free by design. No need to waste 5 seconds at startup any more. DRIVER CHANGES: - Every driver publishes driver up events when starting for the first time or in case of restart when recovery actions must be taken in the upper layers. - Driver up events are published by drivers through DS. - For regular drivers, VFS is normally the only subscriber, but not necessarily. For instance, when the filter driver is in use, it must subscribe to driver up events to initiate recovery. - For network drivers, inet is the only subscriber for now. - Every VFS driver is statically linked with libdriver, every network driver is statically linked with libnetdriver. DRIVER LIBRARIES CHANGES: - Libdriver is extended to provide generic receive() and ds_publish() interfaces for VFS drivers. - driver_receive() is a wrapper for sef_receive() also used in driver_task() to discard spurious messages that were meant to be delivered to a previous version of the driver. - driver_receive_mq() is the same as driver_receive() but integrates support for queued messages. - driver_announce() publishes a driver up event for VFS drivers and marks the driver as initialized and expecting a DEV_OPEN message. - Libnetdriver is introduced to provide similar receive() and ds_publish() interfaces for network drivers (netdriver_announce() and netdriver_receive()). - Network drivers all support live update with no state transfer now. KERNEL CHANGES: - Added kernel call statectl for state management. Used by driver_announce() to unblock eventual callers sendrecing to the driver.
2010-04-08 15:41:35 +02:00
#include <sys/svrctl.h>
2005-04-21 16:53:53 +02:00
Basic System Event Framework (SEF) with ping and live update. SYSLIB CHANGES: - SEF must be used by every system process and is thereby part of the system library. - The framework provides a receive() interface (sef_receive) for system processes to automatically catch known system even messages and process them. - SEF provides a default behavior for each type of system event, but allows system processes to register callbacks to override the default behavior. - Custom (local to the process) or predefined (provided by SEF) callback implementations can be registered to SEF. - SEF currently includes support for 2 types of system events: 1. SEF Ping. The event occurs every time RS sends a ping to figure out whether a system process is still alive. The default callback implementation provided by SEF is to notify RS back to let it know the process is alive and kicking. 2. SEF Live update. The event occurs every time RS sends a prepare to update message to let a system process know an update is available and to prepare for it. The live update support is very basic for now. SEF only deals with verifying if the prepare state can be supported by the process, dumping the state for debugging purposes, and providing an event-driven programming model to the process to react to state changes check-in when ready to update. - SEF should be extended in the future to integrate support for more types of system events. Ideally, all the cross-cutting concerns should be integrated into SEF to avoid duplicating code and ease extensibility. Examples include: * PM notify messages primarily used at shutdown. * SYSTEM notify messages primarily used for signals. * CLOCK notify messages used for system alarms. * Debug messages. IS could still be in charge of fkey handling but would forward the debug message to the target process (e.g. PM, if the user requested debug information about PM). SEF would then catch the message and do nothing unless the process has registered an appropriate callback to deal with the event. This simplifies the programming model to print debug information, avoids duplicating code, and reduces the effort to print debug information. SYSTEM PROCESSES CHANGES: - Every system process registers SEF callbacks it needs to override the default system behavior and calls sef_startup() right after being started. - sef_startup() does almost nothing now, but will be extended in the future to support callbacks of its own to let RS control and synchronize with every system process at initialization time. - Every system process calls sef_receive() now rather than receive() directly, to let SEF handle predefined system events. RS CHANGES: - RS supports a basic single-component live update protocol now, as follows: * When an update command is issued (via "service update *"), RS notifies the target system process to prepare for a specific update state. * If the process doesn't respond back in time, the update is aborted. * When the process responds back, RS kills it and marks it for refreshing. * The process is then automatically restarted as for a buggy process and can start running again. * Live update is currently prototyped as a controlled failure.
2009-12-21 15:12:21 +01:00
/* Variables. */
2005-04-21 16:53:53 +02:00
/* Common command block */
struct command {
u8_t precomp; /* REG_PRECOMP, etc. */
u8_t count;
u8_t sector;
u8_t cyl_lo;
u8_t cyl_hi;
u8_t ldh;
u8_t command;
2006-02-24 13:55:47 +01:00
/* The following at for LBA48 */
u8_t count_prev;
u8_t sector_prev;
u8_t cyl_lo_prev;
u8_t cyl_hi_prev;
2005-04-21 16:53:53 +02:00
};
/* Timeouts and max retries. */
2012-03-25 20:25:53 +02:00
static int timeout_usecs = DEF_TIMEOUT_USECS;
static int max_errors = MAX_ERRORS;
static long w_standard_timeouts = 0;
static long w_pci_debug = 0;
static long w_instance = 0;
static long disable_dma = 0;
static long atapi_debug = 0;
static long w_identify_wakeup_ticks;
static long wakeup_ticks;
static long w_atapi_dma;
2012-03-25 20:25:53 +02:00
static int w_testing = 0;
static int w_silent = 0;
2005-04-21 16:53:53 +02:00
2012-03-25 20:25:53 +02:00
static u32_t system_hz;
/* The struct wini is indexed by drive (0-3). */
2012-03-25 20:25:53 +02:00
static struct wini { /* main drive struct, one entry per drive */
2005-04-21 16:53:53 +02:00
unsigned state; /* drive state: deaf, initialized, dead */
2006-02-24 13:55:47 +01:00
unsigned short w_status; /* device status register */
2005-08-25 18:25:19 +02:00
unsigned base_cmd; /* command base register */
unsigned base_ctl; /* control base register */
2006-03-03 16:37:51 +01:00
unsigned base_dma; /* dma base register */
unsigned char native; /* if set, drive is native (not compat.) */
unsigned char lba48; /* if set, drive supports lba48 */
unsigned char dma; /* if set, drive supports dma */
unsigned char dma_intseen; /* if set, drive has seen an interrupt */
2005-05-02 16:30:04 +02:00
int irq_hook_id; /* id of irq hook at the kernel */
unsigned cylinders; /* physical number of cylinders */
unsigned heads; /* physical number of heads */
unsigned sectors; /* physical number of sectors per track */
2005-04-21 16:53:53 +02:00
unsigned ldhpref; /* top four bytes of the LDH (head) register */
unsigned max_count; /* max request for this drive */
unsigned open_ct; /* in-use count */
struct device part[DEV_PER_DRIVE]; /* disks and partitions */
struct device subpart[SUB_PER_DRIVE]; /* subpartitions */
} wini[MAX_DRIVES], *w_wn;
2012-03-25 20:25:53 +02:00
static int w_device = -1;
2012-03-25 20:25:53 +02:00
int w_command; /* current command in execution */
static int w_drive; /* selected drive */
static struct device *w_dv; /* device's base and size */
2005-04-21 16:53:53 +02:00
2012-03-25 20:25:53 +02:00
static u8_t *tmp_buf;
2006-03-06 11:42:51 +01:00
#define ATA_DMA_SECTORS 64
#define ATA_DMA_BUF_SIZE (ATA_DMA_SECTORS*SECTOR_SIZE)
2006-03-03 16:37:51 +01:00
2012-03-25 20:25:53 +02:00
static char *dma_buf;
static phys_bytes dma_buf_phys;
2006-03-03 16:37:51 +01:00
#define N_PRDTE 1024 /* Should be enough for large requests */
struct prdte
2006-03-03 16:37:51 +01:00
{
phys_bytes prdte_base;
2006-03-03 16:37:51 +01:00
u16_t prdte_count;
u8_t prdte_reserved;
u8_t prdte_flags;
};
#define PRDT_BYTES (sizeof(struct prdte) * N_PRDTE)
2012-03-25 20:25:53 +02:00
static struct prdte *prdt;
static phys_bytes prdt_phys;
2006-03-03 16:37:51 +01:00
#define PRDTE_FL_EOT 0x80 /* End of table */
static int w_probe(int skip, u16_t *vidp, u16_t *didp);
static void w_init(int devind, u16_t vid, u16_t did);
static int init_params(void);
static int w_do_open(devminor_t minor, int access);
static struct device *w_prepare(devminor_t dev);
static struct device *w_part(devminor_t minor);
2012-03-25 20:25:53 +02:00
static int w_identify(void);
static char *w_name(void);
static int w_specify(void);
static int w_io_test(void);
static ssize_t w_transfer(devminor_t minor, int do_write, u64_t position,
endpoint_t proc_nr, iovec_t *iov, unsigned int nr_req, int flags);
2012-03-25 20:25:53 +02:00
static int com_out(struct command *cmd);
static int com_out_ext(struct command *cmd);
static int setup_dma(unsigned *sizep, endpoint_t proc_nr, iovec_t *iov,
size_t addr_offset, int do_write);
2012-03-25 20:25:53 +02:00
static void w_need_reset(void);
static int w_do_close(devminor_t minor);
static int w_ioctl(devminor_t minor, unsigned long request, endpoint_t endpt,
cp_grant_id_t grant, endpoint_t user_endpt);
2012-03-25 20:25:53 +02:00
static void w_hw_int(unsigned int irqs);
static int com_simple(struct command *cmd);
static void w_timeout(void);
static int w_reset(void);
static void w_intr_wait(void);
static int at_intr_wait(void);
static int w_waitfor(int mask, int value);
static int w_waitfor_dma(unsigned int mask, unsigned int value);
static void w_geometry(devminor_t minor, struct part_geom *entry);
2012-03-25 20:25:53 +02:00
static int atapi_sendpacket(u8_t *packet, unsigned cnt, int do_dma);
static int atapi_intr_wait(int dma, size_t max);
static int atapi_open(void);
static void atapi_close(void);
static int atapi_transfer(int do_write, u64_t position, endpoint_t
endpt, iovec_t *iov, unsigned int nr_req);
2005-04-21 16:53:53 +02:00
/* Entry points to this driver. */
2012-03-25 20:25:53 +02:00
static struct blockdriver w_dtab = {
.bdr_type = BLOCKDRIVER_TYPE_DISK, /* handle partition requests */
.bdr_open = w_do_open, /* open or mount request, initialize device */
.bdr_close = w_do_close, /* release device */
.bdr_transfer = w_transfer, /* do the I/O */
.bdr_ioctl = w_ioctl, /* I/O control requests */
.bdr_part = w_part, /* return partition information */
.bdr_geometry = w_geometry, /* tell the geometry of the disk */
.bdr_intr = w_hw_int, /* leftover hardware interrupts */
2005-04-21 16:53:53 +02:00
};
Basic System Event Framework (SEF) with ping and live update. SYSLIB CHANGES: - SEF must be used by every system process and is thereby part of the system library. - The framework provides a receive() interface (sef_receive) for system processes to automatically catch known system even messages and process them. - SEF provides a default behavior for each type of system event, but allows system processes to register callbacks to override the default behavior. - Custom (local to the process) or predefined (provided by SEF) callback implementations can be registered to SEF. - SEF currently includes support for 2 types of system events: 1. SEF Ping. The event occurs every time RS sends a ping to figure out whether a system process is still alive. The default callback implementation provided by SEF is to notify RS back to let it know the process is alive and kicking. 2. SEF Live update. The event occurs every time RS sends a prepare to update message to let a system process know an update is available and to prepare for it. The live update support is very basic for now. SEF only deals with verifying if the prepare state can be supported by the process, dumping the state for debugging purposes, and providing an event-driven programming model to the process to react to state changes check-in when ready to update. - SEF should be extended in the future to integrate support for more types of system events. Ideally, all the cross-cutting concerns should be integrated into SEF to avoid duplicating code and ease extensibility. Examples include: * PM notify messages primarily used at shutdown. * SYSTEM notify messages primarily used for signals. * CLOCK notify messages used for system alarms. * Debug messages. IS could still be in charge of fkey handling but would forward the debug message to the target process (e.g. PM, if the user requested debug information about PM). SEF would then catch the message and do nothing unless the process has registered an appropriate callback to deal with the event. This simplifies the programming model to print debug information, avoids duplicating code, and reduces the effort to print debug information. SYSTEM PROCESSES CHANGES: - Every system process registers SEF callbacks it needs to override the default system behavior and calls sef_startup() right after being started. - sef_startup() does almost nothing now, but will be extended in the future to support callbacks of its own to let RS control and synchronize with every system process at initialization time. - Every system process calls sef_receive() now rather than receive() directly, to let SEF handle predefined system events. RS CHANGES: - RS supports a basic single-component live update protocol now, as follows: * When an update command is issued (via "service update *"), RS notifies the target system process to prepare for a specific update state. * If the process doesn't respond back in time, the update is aborted. * When the process responds back, RS kills it and marks it for refreshing. * The process is then automatically restarted as for a buggy process and can start running again. * Live update is currently prototyped as a controlled failure.
2009-12-21 15:12:21 +01:00
/* SEF functions and variables. */
2012-03-25 20:25:53 +02:00
static void sef_local_startup(void);
static int sef_cb_init_fresh(int type, sef_init_info_t *info);
Basic System Event Framework (SEF) with ping and live update. SYSLIB CHANGES: - SEF must be used by every system process and is thereby part of the system library. - The framework provides a receive() interface (sef_receive) for system processes to automatically catch known system even messages and process them. - SEF provides a default behavior for each type of system event, but allows system processes to register callbacks to override the default behavior. - Custom (local to the process) or predefined (provided by SEF) callback implementations can be registered to SEF. - SEF currently includes support for 2 types of system events: 1. SEF Ping. The event occurs every time RS sends a ping to figure out whether a system process is still alive. The default callback implementation provided by SEF is to notify RS back to let it know the process is alive and kicking. 2. SEF Live update. The event occurs every time RS sends a prepare to update message to let a system process know an update is available and to prepare for it. The live update support is very basic for now. SEF only deals with verifying if the prepare state can be supported by the process, dumping the state for debugging purposes, and providing an event-driven programming model to the process to react to state changes check-in when ready to update. - SEF should be extended in the future to integrate support for more types of system events. Ideally, all the cross-cutting concerns should be integrated into SEF to avoid duplicating code and ease extensibility. Examples include: * PM notify messages primarily used at shutdown. * SYSTEM notify messages primarily used for signals. * CLOCK notify messages used for system alarms. * Debug messages. IS could still be in charge of fkey handling but would forward the debug message to the target process (e.g. PM, if the user requested debug information about PM). SEF would then catch the message and do nothing unless the process has registered an appropriate callback to deal with the event. This simplifies the programming model to print debug information, avoids duplicating code, and reduces the effort to print debug information. SYSTEM PROCESSES CHANGES: - Every system process registers SEF callbacks it needs to override the default system behavior and calls sef_startup() right after being started. - sef_startup() does almost nothing now, but will be extended in the future to support callbacks of its own to let RS control and synchronize with every system process at initialization time. - Every system process calls sef_receive() now rather than receive() directly, to let SEF handle predefined system events. RS CHANGES: - RS supports a basic single-component live update protocol now, as follows: * When an update command is issued (via "service update *"), RS notifies the target system process to prepare for a specific update state. * If the process doesn't respond back in time, the update is aborted. * When the process responds back, RS kills it and marks it for refreshing. * The process is then automatically restarted as for a buggy process and can start running again. * Live update is currently prototyped as a controlled failure.
2009-12-21 15:12:21 +01:00
2005-04-21 16:53:53 +02:00
/*===========================================================================*
* at_winchester_task *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
int main(int argc, char *argv[])
2005-04-21 16:53:53 +02:00
{
Basic System Event Framework (SEF) with ping and live update. SYSLIB CHANGES: - SEF must be used by every system process and is thereby part of the system library. - The framework provides a receive() interface (sef_receive) for system processes to automatically catch known system even messages and process them. - SEF provides a default behavior for each type of system event, but allows system processes to register callbacks to override the default behavior. - Custom (local to the process) or predefined (provided by SEF) callback implementations can be registered to SEF. - SEF currently includes support for 2 types of system events: 1. SEF Ping. The event occurs every time RS sends a ping to figure out whether a system process is still alive. The default callback implementation provided by SEF is to notify RS back to let it know the process is alive and kicking. 2. SEF Live update. The event occurs every time RS sends a prepare to update message to let a system process know an update is available and to prepare for it. The live update support is very basic for now. SEF only deals with verifying if the prepare state can be supported by the process, dumping the state for debugging purposes, and providing an event-driven programming model to the process to react to state changes check-in when ready to update. - SEF should be extended in the future to integrate support for more types of system events. Ideally, all the cross-cutting concerns should be integrated into SEF to avoid duplicating code and ease extensibility. Examples include: * PM notify messages primarily used at shutdown. * SYSTEM notify messages primarily used for signals. * CLOCK notify messages used for system alarms. * Debug messages. IS could still be in charge of fkey handling but would forward the debug message to the target process (e.g. PM, if the user requested debug information about PM). SEF would then catch the message and do nothing unless the process has registered an appropriate callback to deal with the event. This simplifies the programming model to print debug information, avoids duplicating code, and reduces the effort to print debug information. SYSTEM PROCESSES CHANGES: - Every system process registers SEF callbacks it needs to override the default system behavior and calls sef_startup() right after being started. - sef_startup() does almost nothing now, but will be extended in the future to support callbacks of its own to let RS control and synchronize with every system process at initialization time. - Every system process calls sef_receive() now rather than receive() directly, to let SEF handle predefined system events. RS CHANGES: - RS supports a basic single-component live update protocol now, as follows: * When an update command is issued (via "service update *"), RS notifies the target system process to prepare for a specific update state. * If the process doesn't respond back in time, the update is aborted. * When the process responds back, RS kills it and marks it for refreshing. * The process is then automatically restarted as for a buggy process and can start running again. * Live update is currently prototyped as a controlled failure.
2009-12-21 15:12:21 +01:00
/* SEF local startup. */
Initialization protocol for system services. SYSLIB CHANGES: - SEF framework now supports a new SEF Init request type from RS. 3 different callbacks are available (init_fresh, init_lu, init_restart) to specify initialization code when a service starts fresh, starts after a live update, or restarts. SYSTEM SERVICE CHANGES: - Initialization code for system services is now enclosed in a callback SEF will automatically call at init time. The return code of the callback will tell RS whether the initialization completed successfully. - Each init callback can access information passed by RS to initialize. As of now, each system service has access to the public entries of RS's system process table to gather all the information required to initialize. This design eliminates many existing or potential races at boot time and provides a uniform initialization interface to system services. The same interface will be reused for the upcoming publish/subscribe model to handle dynamic registration / deregistration of system services. VM CHANGES: - Uniform privilege management for all system services. Every service uses the same call mask format. For boot services, VM copies the call mask from init data. For dynamic services, VM still receives the call mask via rs_set_priv call that will be soon replaced by the upcoming publish/subscribe model. RS CHANGES: - The system process table has been reorganized and split into private entries and public entries. Only the latter ones are exposed to system services. - VM call masks are now entirely configured in rs/table.c - RS has now its own slot in the system process table. Only kernel tasks and user processes not included in the boot image are now left out from the system process table. - RS implements the initialization protocol for system services. - For services in the boot image, RS blocks till initialization is complete and panics when failure is reported back. Services are initialized in their order of appearance in the boot image priv table and RS blocks to implements synchronous initialization for every system service having the flag SF_SYNCH_BOOT set. - For services started dynamically, the initialization protocol is implemented as though it were the first ping for the service. In this case, if the system service fails to report back (or reports failure), RS brings the service down rather than trying to restart it.
2010-01-08 02:20:42 +01:00
env_setargs(argc, argv);
Basic System Event Framework (SEF) with ping and live update. SYSLIB CHANGES: - SEF must be used by every system process and is thereby part of the system library. - The framework provides a receive() interface (sef_receive) for system processes to automatically catch known system even messages and process them. - SEF provides a default behavior for each type of system event, but allows system processes to register callbacks to override the default behavior. - Custom (local to the process) or predefined (provided by SEF) callback implementations can be registered to SEF. - SEF currently includes support for 2 types of system events: 1. SEF Ping. The event occurs every time RS sends a ping to figure out whether a system process is still alive. The default callback implementation provided by SEF is to notify RS back to let it know the process is alive and kicking. 2. SEF Live update. The event occurs every time RS sends a prepare to update message to let a system process know an update is available and to prepare for it. The live update support is very basic for now. SEF only deals with verifying if the prepare state can be supported by the process, dumping the state for debugging purposes, and providing an event-driven programming model to the process to react to state changes check-in when ready to update. - SEF should be extended in the future to integrate support for more types of system events. Ideally, all the cross-cutting concerns should be integrated into SEF to avoid duplicating code and ease extensibility. Examples include: * PM notify messages primarily used at shutdown. * SYSTEM notify messages primarily used for signals. * CLOCK notify messages used for system alarms. * Debug messages. IS could still be in charge of fkey handling but would forward the debug message to the target process (e.g. PM, if the user requested debug information about PM). SEF would then catch the message and do nothing unless the process has registered an appropriate callback to deal with the event. This simplifies the programming model to print debug information, avoids duplicating code, and reduces the effort to print debug information. SYSTEM PROCESSES CHANGES: - Every system process registers SEF callbacks it needs to override the default system behavior and calls sef_startup() right after being started. - sef_startup() does almost nothing now, but will be extended in the future to support callbacks of its own to let RS control and synchronize with every system process at initialization time. - Every system process calls sef_receive() now rather than receive() directly, to let SEF handle predefined system events. RS CHANGES: - RS supports a basic single-component live update protocol now, as follows: * When an update command is issued (via "service update *"), RS notifies the target system process to prepare for a specific update state. * If the process doesn't respond back in time, the update is aborted. * When the process responds back, RS kills it and marks it for refreshing. * The process is then automatically restarted as for a buggy process and can start running again. * Live update is currently prototyped as a controlled failure.
2009-12-21 15:12:21 +01:00
sef_local_startup();
Initialization protocol for system services. SYSLIB CHANGES: - SEF framework now supports a new SEF Init request type from RS. 3 different callbacks are available (init_fresh, init_lu, init_restart) to specify initialization code when a service starts fresh, starts after a live update, or restarts. SYSTEM SERVICE CHANGES: - Initialization code for system services is now enclosed in a callback SEF will automatically call at init time. The return code of the callback will tell RS whether the initialization completed successfully. - Each init callback can access information passed by RS to initialize. As of now, each system service has access to the public entries of RS's system process table to gather all the information required to initialize. This design eliminates many existing or potential races at boot time and provides a uniform initialization interface to system services. The same interface will be reused for the upcoming publish/subscribe model to handle dynamic registration / deregistration of system services. VM CHANGES: - Uniform privilege management for all system services. Every service uses the same call mask format. For boot services, VM copies the call mask from init data. For dynamic services, VM still receives the call mask via rs_set_priv call that will be soon replaced by the upcoming publish/subscribe model. RS CHANGES: - The system process table has been reorganized and split into private entries and public entries. Only the latter ones are exposed to system services. - VM call masks are now entirely configured in rs/table.c - RS has now its own slot in the system process table. Only kernel tasks and user processes not included in the boot image are now left out from the system process table. - RS implements the initialization protocol for system services. - For services in the boot image, RS blocks till initialization is complete and panics when failure is reported back. Services are initialized in their order of appearance in the boot image priv table and RS blocks to implements synchronous initialization for every system service having the flag SF_SYNCH_BOOT set. - For services started dynamically, the initialization protocol is implemented as though it were the first ping for the service. In this case, if the system service fails to report back (or reports failure), RS brings the service down rather than trying to restart it.
2010-01-08 02:20:42 +01:00
/* Call the generic receive loop. */
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
blockdriver_task(&w_dtab);
Initialization protocol for system services. SYSLIB CHANGES: - SEF framework now supports a new SEF Init request type from RS. 3 different callbacks are available (init_fresh, init_lu, init_restart) to specify initialization code when a service starts fresh, starts after a live update, or restarts. SYSTEM SERVICE CHANGES: - Initialization code for system services is now enclosed in a callback SEF will automatically call at init time. The return code of the callback will tell RS whether the initialization completed successfully. - Each init callback can access information passed by RS to initialize. As of now, each system service has access to the public entries of RS's system process table to gather all the information required to initialize. This design eliminates many existing or potential races at boot time and provides a uniform initialization interface to system services. The same interface will be reused for the upcoming publish/subscribe model to handle dynamic registration / deregistration of system services. VM CHANGES: - Uniform privilege management for all system services. Every service uses the same call mask format. For boot services, VM copies the call mask from init data. For dynamic services, VM still receives the call mask via rs_set_priv call that will be soon replaced by the upcoming publish/subscribe model. RS CHANGES: - The system process table has been reorganized and split into private entries and public entries. Only the latter ones are exposed to system services. - VM call masks are now entirely configured in rs/table.c - RS has now its own slot in the system process table. Only kernel tasks and user processes not included in the boot image are now left out from the system process table. - RS implements the initialization protocol for system services. - For services in the boot image, RS blocks till initialization is complete and panics when failure is reported back. Services are initialized in their order of appearance in the boot image priv table and RS blocks to implements synchronous initialization for every system service having the flag SF_SYNCH_BOOT set. - For services started dynamically, the initialization protocol is implemented as though it were the first ping for the service. In this case, if the system service fails to report back (or reports failure), RS brings the service down rather than trying to restart it.
2010-01-08 02:20:42 +01:00
2005-08-25 15:14:02 +02:00
return(OK);
2005-04-21 16:53:53 +02:00
}
Basic System Event Framework (SEF) with ping and live update. SYSLIB CHANGES: - SEF must be used by every system process and is thereby part of the system library. - The framework provides a receive() interface (sef_receive) for system processes to automatically catch known system even messages and process them. - SEF provides a default behavior for each type of system event, but allows system processes to register callbacks to override the default behavior. - Custom (local to the process) or predefined (provided by SEF) callback implementations can be registered to SEF. - SEF currently includes support for 2 types of system events: 1. SEF Ping. The event occurs every time RS sends a ping to figure out whether a system process is still alive. The default callback implementation provided by SEF is to notify RS back to let it know the process is alive and kicking. 2. SEF Live update. The event occurs every time RS sends a prepare to update message to let a system process know an update is available and to prepare for it. The live update support is very basic for now. SEF only deals with verifying if the prepare state can be supported by the process, dumping the state for debugging purposes, and providing an event-driven programming model to the process to react to state changes check-in when ready to update. - SEF should be extended in the future to integrate support for more types of system events. Ideally, all the cross-cutting concerns should be integrated into SEF to avoid duplicating code and ease extensibility. Examples include: * PM notify messages primarily used at shutdown. * SYSTEM notify messages primarily used for signals. * CLOCK notify messages used for system alarms. * Debug messages. IS could still be in charge of fkey handling but would forward the debug message to the target process (e.g. PM, if the user requested debug information about PM). SEF would then catch the message and do nothing unless the process has registered an appropriate callback to deal with the event. This simplifies the programming model to print debug information, avoids duplicating code, and reduces the effort to print debug information. SYSTEM PROCESSES CHANGES: - Every system process registers SEF callbacks it needs to override the default system behavior and calls sef_startup() right after being started. - sef_startup() does almost nothing now, but will be extended in the future to support callbacks of its own to let RS control and synchronize with every system process at initialization time. - Every system process calls sef_receive() now rather than receive() directly, to let SEF handle predefined system events. RS CHANGES: - RS supports a basic single-component live update protocol now, as follows: * When an update command is issued (via "service update *"), RS notifies the target system process to prepare for a specific update state. * If the process doesn't respond back in time, the update is aborted. * When the process responds back, RS kills it and marks it for refreshing. * The process is then automatically restarted as for a buggy process and can start running again. * Live update is currently prototyped as a controlled failure.
2009-12-21 15:12:21 +01:00
/*===========================================================================*
* sef_local_startup *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static void sef_local_startup(void)
Basic System Event Framework (SEF) with ping and live update. SYSLIB CHANGES: - SEF must be used by every system process and is thereby part of the system library. - The framework provides a receive() interface (sef_receive) for system processes to automatically catch known system even messages and process them. - SEF provides a default behavior for each type of system event, but allows system processes to register callbacks to override the default behavior. - Custom (local to the process) or predefined (provided by SEF) callback implementations can be registered to SEF. - SEF currently includes support for 2 types of system events: 1. SEF Ping. The event occurs every time RS sends a ping to figure out whether a system process is still alive. The default callback implementation provided by SEF is to notify RS back to let it know the process is alive and kicking. 2. SEF Live update. The event occurs every time RS sends a prepare to update message to let a system process know an update is available and to prepare for it. The live update support is very basic for now. SEF only deals with verifying if the prepare state can be supported by the process, dumping the state for debugging purposes, and providing an event-driven programming model to the process to react to state changes check-in when ready to update. - SEF should be extended in the future to integrate support for more types of system events. Ideally, all the cross-cutting concerns should be integrated into SEF to avoid duplicating code and ease extensibility. Examples include: * PM notify messages primarily used at shutdown. * SYSTEM notify messages primarily used for signals. * CLOCK notify messages used for system alarms. * Debug messages. IS could still be in charge of fkey handling but would forward the debug message to the target process (e.g. PM, if the user requested debug information about PM). SEF would then catch the message and do nothing unless the process has registered an appropriate callback to deal with the event. This simplifies the programming model to print debug information, avoids duplicating code, and reduces the effort to print debug information. SYSTEM PROCESSES CHANGES: - Every system process registers SEF callbacks it needs to override the default system behavior and calls sef_startup() right after being started. - sef_startup() does almost nothing now, but will be extended in the future to support callbacks of its own to let RS control and synchronize with every system process at initialization time. - Every system process calls sef_receive() now rather than receive() directly, to let SEF handle predefined system events. RS CHANGES: - RS supports a basic single-component live update protocol now, as follows: * When an update command is issued (via "service update *"), RS notifies the target system process to prepare for a specific update state. * If the process doesn't respond back in time, the update is aborted. * When the process responds back, RS kills it and marks it for refreshing. * The process is then automatically restarted as for a buggy process and can start running again. * Live update is currently prototyped as a controlled failure.
2009-12-21 15:12:21 +01:00
{
Initialization protocol for system services. SYSLIB CHANGES: - SEF framework now supports a new SEF Init request type from RS. 3 different callbacks are available (init_fresh, init_lu, init_restart) to specify initialization code when a service starts fresh, starts after a live update, or restarts. SYSTEM SERVICE CHANGES: - Initialization code for system services is now enclosed in a callback SEF will automatically call at init time. The return code of the callback will tell RS whether the initialization completed successfully. - Each init callback can access information passed by RS to initialize. As of now, each system service has access to the public entries of RS's system process table to gather all the information required to initialize. This design eliminates many existing or potential races at boot time and provides a uniform initialization interface to system services. The same interface will be reused for the upcoming publish/subscribe model to handle dynamic registration / deregistration of system services. VM CHANGES: - Uniform privilege management for all system services. Every service uses the same call mask format. For boot services, VM copies the call mask from init data. For dynamic services, VM still receives the call mask via rs_set_priv call that will be soon replaced by the upcoming publish/subscribe model. RS CHANGES: - The system process table has been reorganized and split into private entries and public entries. Only the latter ones are exposed to system services. - VM call masks are now entirely configured in rs/table.c - RS has now its own slot in the system process table. Only kernel tasks and user processes not included in the boot image are now left out from the system process table. - RS implements the initialization protocol for system services. - For services in the boot image, RS blocks till initialization is complete and panics when failure is reported back. Services are initialized in their order of appearance in the boot image priv table and RS blocks to implements synchronous initialization for every system service having the flag SF_SYNCH_BOOT set. - For services started dynamically, the initialization protocol is implemented as though it were the first ping for the service. In this case, if the system service fails to report back (or reports failure), RS brings the service down rather than trying to restart it.
2010-01-08 02:20:42 +01:00
/* Register init callbacks. */
sef_setcb_init_fresh(sef_cb_init_fresh);
sef_setcb_init_lu(sef_cb_init_fresh);
Basic System Event Framework (SEF) with ping and live update. SYSLIB CHANGES: - SEF must be used by every system process and is thereby part of the system library. - The framework provides a receive() interface (sef_receive) for system processes to automatically catch known system even messages and process them. - SEF provides a default behavior for each type of system event, but allows system processes to register callbacks to override the default behavior. - Custom (local to the process) or predefined (provided by SEF) callback implementations can be registered to SEF. - SEF currently includes support for 2 types of system events: 1. SEF Ping. The event occurs every time RS sends a ping to figure out whether a system process is still alive. The default callback implementation provided by SEF is to notify RS back to let it know the process is alive and kicking. 2. SEF Live update. The event occurs every time RS sends a prepare to update message to let a system process know an update is available and to prepare for it. The live update support is very basic for now. SEF only deals with verifying if the prepare state can be supported by the process, dumping the state for debugging purposes, and providing an event-driven programming model to the process to react to state changes check-in when ready to update. - SEF should be extended in the future to integrate support for more types of system events. Ideally, all the cross-cutting concerns should be integrated into SEF to avoid duplicating code and ease extensibility. Examples include: * PM notify messages primarily used at shutdown. * SYSTEM notify messages primarily used for signals. * CLOCK notify messages used for system alarms. * Debug messages. IS could still be in charge of fkey handling but would forward the debug message to the target process (e.g. PM, if the user requested debug information about PM). SEF would then catch the message and do nothing unless the process has registered an appropriate callback to deal with the event. This simplifies the programming model to print debug information, avoids duplicating code, and reduces the effort to print debug information. SYSTEM PROCESSES CHANGES: - Every system process registers SEF callbacks it needs to override the default system behavior and calls sef_startup() right after being started. - sef_startup() does almost nothing now, but will be extended in the future to support callbacks of its own to let RS control and synchronize with every system process at initialization time. - Every system process calls sef_receive() now rather than receive() directly, to let SEF handle predefined system events. RS CHANGES: - RS supports a basic single-component live update protocol now, as follows: * When an update command is issued (via "service update *"), RS notifies the target system process to prepare for a specific update state. * If the process doesn't respond back in time, the update is aborted. * When the process responds back, RS kills it and marks it for refreshing. * The process is then automatically restarted as for a buggy process and can start running again. * Live update is currently prototyped as a controlled failure.
2009-12-21 15:12:21 +01:00
/* Register live update callbacks. */
sef_setcb_lu_prepare(sef_cb_lu_prepare);
sef_setcb_lu_state_isvalid(sef_cb_lu_state_isvalid);
sef_setcb_lu_state_dump(sef_cb_lu_state_dump);
/* Let SEF perform startup. */
sef_startup();
}
Initialization protocol for system services. SYSLIB CHANGES: - SEF framework now supports a new SEF Init request type from RS. 3 different callbacks are available (init_fresh, init_lu, init_restart) to specify initialization code when a service starts fresh, starts after a live update, or restarts. SYSTEM SERVICE CHANGES: - Initialization code for system services is now enclosed in a callback SEF will automatically call at init time. The return code of the callback will tell RS whether the initialization completed successfully. - Each init callback can access information passed by RS to initialize. As of now, each system service has access to the public entries of RS's system process table to gather all the information required to initialize. This design eliminates many existing or potential races at boot time and provides a uniform initialization interface to system services. The same interface will be reused for the upcoming publish/subscribe model to handle dynamic registration / deregistration of system services. VM CHANGES: - Uniform privilege management for all system services. Every service uses the same call mask format. For boot services, VM copies the call mask from init data. For dynamic services, VM still receives the call mask via rs_set_priv call that will be soon replaced by the upcoming publish/subscribe model. RS CHANGES: - The system process table has been reorganized and split into private entries and public entries. Only the latter ones are exposed to system services. - VM call masks are now entirely configured in rs/table.c - RS has now its own slot in the system process table. Only kernel tasks and user processes not included in the boot image are now left out from the system process table. - RS implements the initialization protocol for system services. - For services in the boot image, RS blocks till initialization is complete and panics when failure is reported back. Services are initialized in their order of appearance in the boot image priv table and RS blocks to implements synchronous initialization for every system service having the flag SF_SYNCH_BOOT set. - For services started dynamically, the initialization protocol is implemented as though it were the first ping for the service. In this case, if the system service fails to report back (or reports failure), RS brings the service down rather than trying to restart it.
2010-01-08 02:20:42 +01:00
/*===========================================================================*
* sef_cb_init_fresh *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static int sef_cb_init_fresh(int type, sef_init_info_t *UNUSED(info))
Initialization protocol for system services. SYSLIB CHANGES: - SEF framework now supports a new SEF Init request type from RS. 3 different callbacks are available (init_fresh, init_lu, init_restart) to specify initialization code when a service starts fresh, starts after a live update, or restarts. SYSTEM SERVICE CHANGES: - Initialization code for system services is now enclosed in a callback SEF will automatically call at init time. The return code of the callback will tell RS whether the initialization completed successfully. - Each init callback can access information passed by RS to initialize. As of now, each system service has access to the public entries of RS's system process table to gather all the information required to initialize. This design eliminates many existing or potential races at boot time and provides a uniform initialization interface to system services. The same interface will be reused for the upcoming publish/subscribe model to handle dynamic registration / deregistration of system services. VM CHANGES: - Uniform privilege management for all system services. Every service uses the same call mask format. For boot services, VM copies the call mask from init data. For dynamic services, VM still receives the call mask via rs_set_priv call that will be soon replaced by the upcoming publish/subscribe model. RS CHANGES: - The system process table has been reorganized and split into private entries and public entries. Only the latter ones are exposed to system services. - VM call masks are now entirely configured in rs/table.c - RS has now its own slot in the system process table. Only kernel tasks and user processes not included in the boot image are now left out from the system process table. - RS implements the initialization protocol for system services. - For services in the boot image, RS blocks till initialization is complete and panics when failure is reported back. Services are initialized in their order of appearance in the boot image priv table and RS blocks to implements synchronous initialization for every system service having the flag SF_SYNCH_BOOT set. - For services started dynamically, the initialization protocol is implemented as though it were the first ping for the service. In this case, if the system service fails to report back (or reports failure), RS brings the service down rather than trying to restart it.
2010-01-08 02:20:42 +01:00
{
/* Initialize the at_wini driver. */
int skip, devind;
u16_t vid, did;
Initialization protocol for system services. SYSLIB CHANGES: - SEF framework now supports a new SEF Init request type from RS. 3 different callbacks are available (init_fresh, init_lu, init_restart) to specify initialization code when a service starts fresh, starts after a live update, or restarts. SYSTEM SERVICE CHANGES: - Initialization code for system services is now enclosed in a callback SEF will automatically call at init time. The return code of the callback will tell RS whether the initialization completed successfully. - Each init callback can access information passed by RS to initialize. As of now, each system service has access to the public entries of RS's system process table to gather all the information required to initialize. This design eliminates many existing or potential races at boot time and provides a uniform initialization interface to system services. The same interface will be reused for the upcoming publish/subscribe model to handle dynamic registration / deregistration of system services. VM CHANGES: - Uniform privilege management for all system services. Every service uses the same call mask format. For boot services, VM copies the call mask from init data. For dynamic services, VM still receives the call mask via rs_set_priv call that will be soon replaced by the upcoming publish/subscribe model. RS CHANGES: - The system process table has been reorganized and split into private entries and public entries. Only the latter ones are exposed to system services. - VM call masks are now entirely configured in rs/table.c - RS has now its own slot in the system process table. Only kernel tasks and user processes not included in the boot image are now left out from the system process table. - RS implements the initialization protocol for system services. - For services in the boot image, RS blocks till initialization is complete and panics when failure is reported back. Services are initialized in their order of appearance in the boot image priv table and RS blocks to implements synchronous initialization for every system service having the flag SF_SYNCH_BOOT set. - For services started dynamically, the initialization protocol is implemented as though it were the first ping for the service. In this case, if the system service fails to report back (or reports failure), RS brings the service down rather than trying to restart it.
2010-01-08 02:20:42 +01:00
system_hz = sys_hz();
if (!(tmp_buf = alloc_contig(2*DMA_BUF_SIZE, AC_ALIGN4K, NULL)))
panic("unable to allocate temporary buffer");
Initialization protocol for system services. SYSLIB CHANGES: - SEF framework now supports a new SEF Init request type from RS. 3 different callbacks are available (init_fresh, init_lu, init_restart) to specify initialization code when a service starts fresh, starts after a live update, or restarts. SYSTEM SERVICE CHANGES: - Initialization code for system services is now enclosed in a callback SEF will automatically call at init time. The return code of the callback will tell RS whether the initialization completed successfully. - Each init callback can access information passed by RS to initialize. As of now, each system service has access to the public entries of RS's system process table to gather all the information required to initialize. This design eliminates many existing or potential races at boot time and provides a uniform initialization interface to system services. The same interface will be reused for the upcoming publish/subscribe model to handle dynamic registration / deregistration of system services. VM CHANGES: - Uniform privilege management for all system services. Every service uses the same call mask format. For boot services, VM copies the call mask from init data. For dynamic services, VM still receives the call mask via rs_set_priv call that will be soon replaced by the upcoming publish/subscribe model. RS CHANGES: - The system process table has been reorganized and split into private entries and public entries. Only the latter ones are exposed to system services. - VM call masks are now entirely configured in rs/table.c - RS has now its own slot in the system process table. Only kernel tasks and user processes not included in the boot image are now left out from the system process table. - RS implements the initialization protocol for system services. - For services in the boot image, RS blocks till initialization is complete and panics when failure is reported back. Services are initialized in their order of appearance in the boot image priv table and RS blocks to implements synchronous initialization for every system service having the flag SF_SYNCH_BOOT set. - For services started dynamically, the initialization protocol is implemented as though it were the first ping for the service. In this case, if the system service fails to report back (or reports failure), RS brings the service down rather than trying to restart it.
2010-01-08 02:20:42 +01:00
w_identify_wakeup_ticks = WAKEUP_TICKS;
wakeup_ticks = WAKEUP_TICKS;
/* Set special disk parameters. */
skip = init_params();
/* Find the PCI device to use. If none found, terminate immediately. */
devind = w_probe(skip, &vid, &did);
if (devind < 0) {
/* For now, complain only if even the first at_wini instance cannot
* find a device. There may be only one IDE controller after all,
* but if there are none, the system should probably be booted with
* another driver, and that's something the user might want to know.
*/
if (w_instance == 0)
panic("no matching device found");
return ENODEV; /* the actual error code doesn't matter */
}
/* Initialize the device. */
w_init(devind, vid, did);
Initialization protocol for system services. SYSLIB CHANGES: - SEF framework now supports a new SEF Init request type from RS. 3 different callbacks are available (init_fresh, init_lu, init_restart) to specify initialization code when a service starts fresh, starts after a live update, or restarts. SYSTEM SERVICE CHANGES: - Initialization code for system services is now enclosed in a callback SEF will automatically call at init time. The return code of the callback will tell RS whether the initialization completed successfully. - Each init callback can access information passed by RS to initialize. As of now, each system service has access to the public entries of RS's system process table to gather all the information required to initialize. This design eliminates many existing or potential races at boot time and provides a uniform initialization interface to system services. The same interface will be reused for the upcoming publish/subscribe model to handle dynamic registration / deregistration of system services. VM CHANGES: - Uniform privilege management for all system services. Every service uses the same call mask format. For boot services, VM copies the call mask from init data. For dynamic services, VM still receives the call mask via rs_set_priv call that will be soon replaced by the upcoming publish/subscribe model. RS CHANGES: - The system process table has been reorganized and split into private entries and public entries. Only the latter ones are exposed to system services. - VM call masks are now entirely configured in rs/table.c - RS has now its own slot in the system process table. Only kernel tasks and user processes not included in the boot image are now left out from the system process table. - RS implements the initialization protocol for system services. - For services in the boot image, RS blocks till initialization is complete and panics when failure is reported back. Services are initialized in their order of appearance in the boot image priv table and RS blocks to implements synchronous initialization for every system service having the flag SF_SYNCH_BOOT set. - For services started dynamically, the initialization protocol is implemented as though it were the first ping for the service. In this case, if the system service fails to report back (or reports failure), RS brings the service down rather than trying to restart it.
2010-01-08 02:20:42 +01:00
Driver refactory for live update and crash recovery. SYSLIB CHANGES: - DS calls to publish / retrieve labels consider endpoints instead of u32_t. VFS CHANGES: - mapdriver() only adds an entry in the dmap table in VFS. - dev_up() is only executed upon reception of a driver up event. INET CHANGES: - INET no longer searches for existing drivers instances at startup. - A newtwork driver is (re)initialized upon reception of a driver up event. - Networking startup is now race-free by design. No need to waste 5 seconds at startup any more. DRIVER CHANGES: - Every driver publishes driver up events when starting for the first time or in case of restart when recovery actions must be taken in the upper layers. - Driver up events are published by drivers through DS. - For regular drivers, VFS is normally the only subscriber, but not necessarily. For instance, when the filter driver is in use, it must subscribe to driver up events to initiate recovery. - For network drivers, inet is the only subscriber for now. - Every VFS driver is statically linked with libdriver, every network driver is statically linked with libnetdriver. DRIVER LIBRARIES CHANGES: - Libdriver is extended to provide generic receive() and ds_publish() interfaces for VFS drivers. - driver_receive() is a wrapper for sef_receive() also used in driver_task() to discard spurious messages that were meant to be delivered to a previous version of the driver. - driver_receive_mq() is the same as driver_receive() but integrates support for queued messages. - driver_announce() publishes a driver up event for VFS drivers and marks the driver as initialized and expecting a DEV_OPEN message. - Libnetdriver is introduced to provide similar receive() and ds_publish() interfaces for network drivers (netdriver_announce() and netdriver_receive()). - Network drivers all support live update with no state transfer now. KERNEL CHANGES: - Added kernel call statectl for state management. Used by driver_announce() to unblock eventual callers sendrecing to the driver.
2010-04-08 15:41:35 +02:00
/* Announce we are up! */
blockdriver_announce(type);
Driver refactory for live update and crash recovery. SYSLIB CHANGES: - DS calls to publish / retrieve labels consider endpoints instead of u32_t. VFS CHANGES: - mapdriver() only adds an entry in the dmap table in VFS. - dev_up() is only executed upon reception of a driver up event. INET CHANGES: - INET no longer searches for existing drivers instances at startup. - A newtwork driver is (re)initialized upon reception of a driver up event. - Networking startup is now race-free by design. No need to waste 5 seconds at startup any more. DRIVER CHANGES: - Every driver publishes driver up events when starting for the first time or in case of restart when recovery actions must be taken in the upper layers. - Driver up events are published by drivers through DS. - For regular drivers, VFS is normally the only subscriber, but not necessarily. For instance, when the filter driver is in use, it must subscribe to driver up events to initiate recovery. - For network drivers, inet is the only subscriber for now. - Every VFS driver is statically linked with libdriver, every network driver is statically linked with libnetdriver. DRIVER LIBRARIES CHANGES: - Libdriver is extended to provide generic receive() and ds_publish() interfaces for VFS drivers. - driver_receive() is a wrapper for sef_receive() also used in driver_task() to discard spurious messages that were meant to be delivered to a previous version of the driver. - driver_receive_mq() is the same as driver_receive() but integrates support for queued messages. - driver_announce() publishes a driver up event for VFS drivers and marks the driver as initialized and expecting a DEV_OPEN message. - Libnetdriver is introduced to provide similar receive() and ds_publish() interfaces for network drivers (netdriver_announce() and netdriver_receive()). - Network drivers all support live update with no state transfer now. KERNEL CHANGES: - Added kernel call statectl for state management. Used by driver_announce() to unblock eventual callers sendrecing to the driver.
2010-04-08 15:41:35 +02:00
Initialization protocol for system services. SYSLIB CHANGES: - SEF framework now supports a new SEF Init request type from RS. 3 different callbacks are available (init_fresh, init_lu, init_restart) to specify initialization code when a service starts fresh, starts after a live update, or restarts. SYSTEM SERVICE CHANGES: - Initialization code for system services is now enclosed in a callback SEF will automatically call at init time. The return code of the callback will tell RS whether the initialization completed successfully. - Each init callback can access information passed by RS to initialize. As of now, each system service has access to the public entries of RS's system process table to gather all the information required to initialize. This design eliminates many existing or potential races at boot time and provides a uniform initialization interface to system services. The same interface will be reused for the upcoming publish/subscribe model to handle dynamic registration / deregistration of system services. VM CHANGES: - Uniform privilege management for all system services. Every service uses the same call mask format. For boot services, VM copies the call mask from init data. For dynamic services, VM still receives the call mask via rs_set_priv call that will be soon replaced by the upcoming publish/subscribe model. RS CHANGES: - The system process table has been reorganized and split into private entries and public entries. Only the latter ones are exposed to system services. - VM call masks are now entirely configured in rs/table.c - RS has now its own slot in the system process table. Only kernel tasks and user processes not included in the boot image are now left out from the system process table. - RS implements the initialization protocol for system services. - For services in the boot image, RS blocks till initialization is complete and panics when failure is reported back. Services are initialized in their order of appearance in the boot image priv table and RS blocks to implements synchronous initialization for every system service having the flag SF_SYNCH_BOOT set. - For services started dynamically, the initialization protocol is implemented as though it were the first ping for the service. In this case, if the system service fails to report back (or reports failure), RS brings the service down rather than trying to restart it.
2010-01-08 02:20:42 +01:00
return(OK);
}
2005-09-11 19:09:11 +02:00
/*===========================================================================*
* init_params *
*===========================================================================*/
static int init_params(void)
2005-04-21 16:53:53 +02:00
{
/* This routine is called at startup to initialize the drive parameters. */
int drive;
2007-02-16 16:56:00 +01:00
long wakeup_secs = WAKEUP_SECS;
2005-04-21 16:53:53 +02:00
/* Boot variables. */
env_parse("instance", "d", 0, &w_instance, 0, 8);
env_parse("ata_std_timeout", "d", 0, &w_standard_timeouts, 0, 1);
env_parse("ata_pci_debug", "d", 0, &w_pci_debug, 0, 1);
2009-04-27 13:53:11 +02:00
env_parse(NO_DMA_VAR, "d", 0, &disable_dma, 0, 1);
env_parse("ata_id_timeout", "d", 0, &wakeup_secs, 1, 60);
env_parse("atapi_debug", "d", 0, &atapi_debug, 0, 1);
env_parse("atapi_dma", "d", 0, &w_atapi_dma, 0, 1);
w_identify_wakeup_ticks = wakeup_secs * system_hz;
if(atapi_debug)
panic("atapi_debug");
if(w_identify_wakeup_ticks <= 0) {
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
printf("changing wakeup from %ld to %d ticks.\n",
w_identify_wakeup_ticks, WAKEUP_TICKS);
w_identify_wakeup_ticks = WAKEUP_TICKS;
}
if (disable_dma) {
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
printf("at_wini%ld: DMA for ATA devices is disabled.\n", w_instance);
} else {
/* Ask for anonymous memory for DMA, that is physically contiguous. */
dma_buf = alloc_contig(ATA_DMA_BUF_SIZE, 0, &dma_buf_phys);
prdt = alloc_contig(PRDT_BYTES, 0, &prdt_phys);
if(!dma_buf || !prdt) {
disable_dma = 1;
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
printf("at_wini%ld: no dma\n", w_instance);
}
}
2006-03-03 16:37:51 +01:00
for (drive = 0; drive < MAX_DRIVES; drive++)
wini[drive].state = IGNORING;
return (int) w_instance;
}
2005-09-11 19:09:11 +02:00
/*===========================================================================*
* init_drive *
*===========================================================================*/
static void init_drive(int drive, int base_cmd, int base_ctl, int base_dma,
int native, int hook)
{
struct wini *w;
w = &wini[drive];
w->state = 0;
w->w_status = 0;
w->base_cmd = base_cmd;
w->base_ctl = base_ctl;
w->base_dma = base_dma;
if (w_pci_debug)
printf("at_wini%ld: drive %d: base_cmd 0x%x, base_ctl 0x%x, "
"base_dma 0x%x\n", w_instance, drive, w->base_cmd, w->base_ctl,
w->base_dma);
w->native = native;
w->irq_hook_id = hook;
w->ldhpref = ldh_init(drive);
w->max_count = MAX_SECS << SECTOR_SHIFT;
w->lba48 = 0;
w->dma = 0;
}
/*===========================================================================*
* w_probe *
*===========================================================================*/
static int w_probe(int skip, u16_t *vidp, u16_t *didp)
{
/* Go through the PCI devices that have been made visible to us, skipping as
* many as requested and then reserving the first one after that. We assume
* that all visible devices are in fact devices we can handle.
*/
int r, devind;
pci_init();
r = pci_first_dev(&devind, vidp, didp);
if (r <= 0)
return -1;
while (skip--) {
r = pci_next_dev(&devind, vidp, didp);
if (r <= 0)
return -1;
}
pci_reserve(devind);
return devind;
}
2005-09-11 19:09:11 +02:00
/*===========================================================================*
* w_init *
2005-09-11 19:09:11 +02:00
*===========================================================================*/
static void w_init(int devind, u16_t vid, u16_t did)
{
/* Initialize drives on the controller that we found and reserved. Each
* controller has two channels, each of which may have up to two disks
* attached, so the maximum number of disks per controller is always four.
* In this function, we always initialize the slots for all four disks; later,
* during normal operation, we determine whether the disks are actually there.
* For pure IDE devices (as opposed to e.g. RAID devices), each of the two
* channels on the controller may be in native or compatibility mode. The PCI
* interface field tells us which channel is in which mode. For native
* channels, we get the IRQ and the channel's base control and command
* addresses from the PCI slot, and we manually acknowledge interrupts. For
* compatibility channels, we use the hardcoded legacy IRQs and addresses, and
* enable automatic IRQ reenabling. In both cases, we get the base DMA address
* from the PCI slot if it is there.
*/
int r, irq, native_hook, compat_hook, is_ide, nhooks;
u8_t bcr, scr, interface;
u16_t cr;
u32_t base_cmd, base_ctl, base_dma;
2006-03-03 16:37:51 +01:00
bcr= pci_attr_r8(devind, PCI_BCR);
scr= pci_attr_r8(devind, PCI_SCR);
interface= pci_attr_r8(devind, PCI_PIFR);
2005-12-02 15:45:10 +01:00
is_ide = (bcr == PCI_BCR_MASS_STORAGE && scr == PCI_MS_IDE);
irq = pci_attr_r8(devind, PCI_ILR);
base_dma = pci_attr_r32(devind, PCI_BAR_5) & PCI_BAR_IO_MASK;
2006-03-03 16:37:51 +01:00
nhooks = 0; /* we don't care about notify IDs, but they must be unique */
/* Any native drives? Then register their native IRQ first. */
if (!is_ide || (interface & (ATA_IF_NATIVE0 | ATA_IF_NATIVE1))) {
native_hook = nhooks++;
if ((r = sys_irqsetpolicy(irq, 0, &native_hook)) != OK)
panic("couldn't set native IRQ policy %d: %d", irq, r);
if ((r = sys_irqenable(&native_hook)) != OK)
panic("couldn't enable native IRQ line %d: %d", irq, r);
}
/* Add drives on the primary channel. */
if (!is_ide || (interface & ATA_IF_NATIVE0)) {
base_cmd = pci_attr_r32(devind, PCI_BAR) & PCI_BAR_IO_MASK;
base_ctl = pci_attr_r32(devind, PCI_BAR_2) & PCI_BAR_IO_MASK;
init_drive(0, base_cmd, base_ctl+PCI_CTL_OFF, base_dma, TRUE,
native_hook);
init_drive(1, base_cmd, base_ctl+PCI_CTL_OFF, base_dma, TRUE,
native_hook);
if (w_pci_debug)
printf("at_wini%ld: native 0 on %d: 0x%x 0x%x irq %d\n",
w_instance, devind, base_cmd, base_ctl, irq);
} else {
/* Register first compatibility IRQ. */
compat_hook = nhooks++;
if ((r = sys_irqsetpolicy(AT_WINI_0_IRQ, IRQ_REENABLE,
&compat_hook)) != OK)
panic("couldn't set compat(0) IRQ policy: %d", r);
if ((r = sys_irqenable(&compat_hook)) != OK)
panic("couldn't enable compat(0) IRQ line: %d", r);
init_drive(0, REG_CMD_BASE0, REG_CTL_BASE0, base_dma, FALSE,
compat_hook);
init_drive(1, REG_CMD_BASE0, REG_CTL_BASE0, base_dma, FALSE,
compat_hook);
if (w_pci_debug)
printf("at_wini%ld: compat 0 on %d\n", w_instance, devind);
}
/* Add drives on the secondary channel. */
if (base_dma != 0)
base_dma += PCI_DMA_2ND_OFF;
if (!is_ide || (interface & ATA_IF_NATIVE1)) {
base_cmd = pci_attr_r32(devind, PCI_BAR_3) & PCI_BAR_IO_MASK;
base_ctl = pci_attr_r32(devind, PCI_BAR_4) & PCI_BAR_IO_MASK;
init_drive(2, base_cmd, base_ctl+PCI_CTL_OFF, base_dma, TRUE,
native_hook);
init_drive(3, base_cmd, base_ctl+PCI_CTL_OFF, base_dma, TRUE,
native_hook);
if (w_pci_debug)
printf("at_wini%ld: native 1 on %d: 0x%x 0x%x irq %d\n",
w_instance, devind, base_cmd, base_ctl, irq);
} else {
/* Register secondary compatibility IRQ. */
compat_hook = nhooks++;
if ((r = sys_irqsetpolicy(AT_WINI_1_IRQ, IRQ_REENABLE,
&compat_hook)) != OK)
panic("couldn't set compat(1) IRQ policy: %d", r);
if ((r = sys_irqenable(&compat_hook)) != OK)
panic("couldn't enable compat(1) IRQ line: %d", r);
init_drive(2, REG_CMD_BASE1, REG_CTL_BASE1, base_dma, FALSE,
compat_hook);
init_drive(3, REG_CMD_BASE1, REG_CTL_BASE1, base_dma, FALSE,
compat_hook);
if (w_pci_debug)
printf("at_wini%ld: compat 1 on %d\n", w_instance, devind);
}
/* Enable busmastering if necessary. */
cr = pci_attr_r16(devind, PCI_CR);
if (!(cr & PCI_CR_MAST_EN))
pci_attr_w16(devind, PCI_CR, cr | PCI_CR_MAST_EN);
2005-04-21 16:53:53 +02:00
}
2005-09-11 19:09:11 +02:00
/*===========================================================================*
* w_do_open *
*===========================================================================*/
static int w_do_open(devminor_t minor, int access)
2005-04-21 16:53:53 +02:00
{
/* Device open: Initialize the controller and read the partition table. */
struct wini *wn;
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
if (w_prepare(minor) == NULL) return(ENXIO);
2005-04-21 16:53:53 +02:00
wn = w_wn;
/* If we've probed it before and it failed, don't probe it again. */
if (wn->state & IGNORING) return ENXIO;
/* If we haven't identified it yet, or it's gone deaf,
* (re-)identify it.
*/
if (!(wn->state & IDENTIFIED) || (wn->state & DEAF)) {
2005-04-21 16:53:53 +02:00
/* Try to identify the device. */
if (w_identify() != OK) {
#if VERBOSE
printf("%s: identification failed\n", w_name());
#endif
if (wn->state & DEAF){
int err = w_reset();
if( err != OK ){
return err;
}
}
wn->state = IGNORING;
2005-04-21 16:53:53 +02:00
return(ENXIO);
}
/* Do a test transaction unless it's a CD drive (then
* we can believe the controller, and a test may fail
* due to no CD being in the drive). If it fails, ignore
* the device forever.
*/
2005-09-11 19:09:11 +02:00
if (!(wn->state & ATAPI) && w_io_test() != OK) {
wn->state |= IGNORING;
return(ENXIO);
}
2005-04-21 16:53:53 +02:00
}
if ((wn->state & ATAPI) && (access & BDEV_W_BIT))
return(EACCES);
/* Partition the drive if it's being opened for the first time,
* or being opened after being closed.
*/
2005-04-21 16:53:53 +02:00
if (wn->open_ct == 0) {
if (wn->state & ATAPI) {
int r;
if ((r = atapi_open()) != OK) return(r);
}
2005-04-21 16:53:53 +02:00
/* Partition the disk. */
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
partition(&w_dtab, w_drive * DEV_PER_DRIVE, P_PRIMARY,
wn->state & ATAPI);
2005-04-21 16:53:53 +02:00
}
wn->open_ct++;
2005-04-21 16:53:53 +02:00
return(OK);
}
/*===========================================================================*
* w_prepare *
*===========================================================================*/
static struct device *w_prepare(devminor_t device)
2005-04-21 16:53:53 +02:00
{
/* Prepare for I/O on a device. */
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
w_device = (int) device;
if (device >= 0 && device < NR_MINORS) { /* d0, d0p[0-3], d1, ... */
2005-04-21 16:53:53 +02:00
w_drive = device / DEV_PER_DRIVE; /* save drive number */
if (w_drive >= MAX_DRIVES) return(NULL);
2005-04-21 16:53:53 +02:00
w_wn = &wini[w_drive];
w_dv = &w_wn->part[device % DEV_PER_DRIVE];
} else
if ((unsigned) (device -= MINOR_d0p0s0) < NR_SUBDEVS) {/*d[0-7]p[0-3]s[0-3]*/
w_drive = device / SUB_PER_DRIVE;
if (w_drive >= MAX_DRIVES) return(NULL);
2005-04-21 16:53:53 +02:00
w_wn = &wini[w_drive];
w_dv = &w_wn->subpart[device % SUB_PER_DRIVE];
} else {
w_device = -1;
return(NULL);
2005-04-21 16:53:53 +02:00
}
return(w_dv);
}
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
/*===========================================================================*
* w_part *
*===========================================================================*/
static struct device *w_part(devminor_t device)
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
{
/* Return a pointer to the partition information of the given minor device. */
return w_prepare(device);
}
#define id_byte(n) (&tmp_buf[2 * (n)])
#define id_word(n) (((u16_t) id_byte(n)[0] << 0) \
|((u16_t) id_byte(n)[1] << 8))
#define id_longword(n) (((u32_t) id_byte(n)[0] << 0) \
|((u32_t) id_byte(n)[1] << 8) \
|((u32_t) id_byte(n)[2] << 16) \
|((u32_t) id_byte(n)[3] << 24))
/*===========================================================================*
* check_dma *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static void
check_dma(struct wini *wn)
{
2012-03-05 00:11:41 +01:00
u32_t dma_status, dma_base;
int id_dma, ultra_dma;
u16_t w;
wn->dma= 0;
if (disable_dma)
return;
w= id_word(ID_CAPABILITIES);
id_dma= !!(w & ID_CAP_DMA);
w= id_byte(ID_FIELD_VALIDITY)[0];
ultra_dma= !!(w & ID_FV_88);
dma_base= wn->base_dma;
if (dma_base) {
if (sys_inb(dma_base + DMA_STATUS, &dma_status) != OK) {
panic("unable to read DMA status register");
}
}
if (id_dma && dma_base) {
w= id_word(ID_MULTIWORD_DMA);
if (w_pci_debug &&
(w & (ID_MWDMA_2_SUP|ID_MWDMA_1_SUP|ID_MWDMA_0_SUP))) {
printf(
"%s: multiword DMA modes supported:%s%s%s\n",
w_name(),
(w & ID_MWDMA_0_SUP) ? " 0" : "",
(w & ID_MWDMA_1_SUP) ? " 1" : "",
(w & ID_MWDMA_2_SUP) ? " 2" : "");
}
if (w_pci_debug &&
(w & (ID_MWDMA_0_SEL|ID_MWDMA_1_SEL|ID_MWDMA_2_SEL))) {
printf(
"%s: multiword DMA mode selected:%s%s%s\n",
w_name(),
(w & ID_MWDMA_0_SEL) ? " 0" : "",
(w & ID_MWDMA_1_SEL) ? " 1" : "",
(w & ID_MWDMA_2_SEL) ? " 2" : "");
}
if (w_pci_debug && ultra_dma) {
w= id_word(ID_ULTRA_DMA);
if (w & (ID_UDMA_0_SUP|ID_UDMA_1_SUP|
ID_UDMA_2_SUP|ID_UDMA_3_SUP|
ID_UDMA_4_SUP|ID_UDMA_5_SUP)) {
printf(
"%s: Ultra DMA modes supported:%s%s%s%s%s%s\n",
w_name(),
(w & ID_UDMA_0_SUP) ? " 0" : "",
(w & ID_UDMA_1_SUP) ? " 1" : "",
(w & ID_UDMA_2_SUP) ? " 2" : "",
(w & ID_UDMA_3_SUP) ? " 3" : "",
(w & ID_UDMA_4_SUP) ? " 4" : "",
(w & ID_UDMA_5_SUP) ? " 5" : "");
}
if (w & (ID_UDMA_0_SEL|ID_UDMA_1_SEL|
ID_UDMA_2_SEL|ID_UDMA_3_SEL|
ID_UDMA_4_SEL|ID_UDMA_5_SEL)) {
printf(
"%s: Ultra DMA mode selected:%s%s%s%s%s%s\n",
w_name(),
(w & ID_UDMA_0_SEL) ? " 0" : "",
(w & ID_UDMA_1_SEL) ? " 1" : "",
(w & ID_UDMA_2_SEL) ? " 2" : "",
(w & ID_UDMA_3_SEL) ? " 3" : "",
(w & ID_UDMA_4_SEL) ? " 4" : "",
(w & ID_UDMA_5_SEL) ? " 5" : "");
}
}
wn->dma= 1;
} else if (id_dma || dma_base) {
printf("id_dma %d, dma_base 0x%x\n", id_dma, dma_base);
} else
printf("no DMA support\n");
}
2005-04-21 16:53:53 +02:00
/*===========================================================================*
* w_identify *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static int w_identify(void)
2005-04-21 16:53:53 +02:00
{
/* Find out if a device exists, if it is an old AT disk, or a newer ATA
* drive, a removable media device, etc.
*/
struct wini *wn = w_wn;
struct command cmd;
int s;
2006-02-24 13:55:47 +01:00
u16_t w;
2005-04-21 16:53:53 +02:00
unsigned long size;
int prev_wakeup;
int r;
2005-04-21 16:53:53 +02:00
/* Try to identify the device. */
cmd.ldh = wn->ldhpref;
cmd.command = ATA_IDENTIFY;
/* In testing mode, a drive will get ignored at the first timeout. */
w_testing = 1;
/* Execute *_IDENTIFY with configured *_IDENTIFY timeout. */
prev_wakeup = wakeup_ticks;
wakeup_ticks = w_identify_wakeup_ticks;
r = com_simple(&cmd);
if (r == OK && w_waitfor(STATUS_DRQ, STATUS_DRQ) &&
2006-03-09 16:05:43 +01:00
!(wn->w_status & (STATUS_ERR|STATUS_WF))) {
2005-04-21 16:53:53 +02:00
/* Device information. */
2005-08-25 18:25:19 +02:00
if ((s=sys_insw(wn->base_cmd + REG_DATA, SELF, tmp_buf, SECTOR_SIZE)) != OK)
panic("Call to sys_insw() failed: %d", s);
2005-04-21 16:53:53 +02:00
2006-03-09 16:05:43 +01:00
/* This is an ATA device. */
wn->state |= SMART;
2005-04-21 16:53:53 +02:00
/* Preferred CHS translation mode. */
wn->cylinders = id_word(1);
wn->heads = id_word(3);
wn->sectors = id_word(6);
size = (u32_t) wn->cylinders * wn->heads * wn->sectors;
2005-04-21 16:53:53 +02:00
2006-02-24 13:55:47 +01:00
w= id_word(ID_CAPABILITIES);
if ((w & ID_CAP_LBA) && size > 512L*1024*2) {
2005-04-21 16:53:53 +02:00
/* Drive is LBA capable and is big enough to trust it to
* not make a mess of it.
*/
wn->ldhpref |= LDH_LBA;
size = id_longword(60);
2005-08-25 18:25:19 +02:00
2006-02-24 13:55:47 +01:00
w= id_word(ID_CSS);
if (size < LBA48_CHECK_SIZE)
{
/* No need to check for LBA48 */
}
else if (w & ID_CSS_LBA48) {
2005-08-25 18:25:19 +02:00
/* Drive is LBA48 capable (and LBA48 is turned on). */
2006-02-24 13:55:47 +01:00
if (id_longword(102)) {
2005-08-25 18:25:19 +02:00
/* If no. of sectors doesn't fit in 32 bits,
* trunacte to this. So it's LBA32 for now.
* This can still address devices up to 2TB
* though.
*/
size = ULONG_MAX;
} else {
/* Actual number of sectors fits in 32 bits. */
size = id_longword(100);
}
wn->lba48 = 1;
}
2006-03-03 16:37:51 +01:00
check_dma(wn);
2005-04-21 16:53:53 +02:00
}
} else if (cmd.command = ATAPI_IDENTIFY,
2006-03-09 16:05:43 +01:00
com_simple(&cmd) == OK && w_waitfor(STATUS_DRQ, STATUS_DRQ) &&
!(wn->w_status & (STATUS_ERR|STATUS_WF))) {
2005-04-21 16:53:53 +02:00
/* An ATAPI device. */
wn->state |= ATAPI;
/* Device information. */
2005-08-25 18:25:19 +02:00
if ((s=sys_insw(wn->base_cmd + REG_DATA, SELF, tmp_buf, 512)) != OK)
panic("Call to sys_insw() failed: %d", s);
2005-04-21 16:53:53 +02:00
size = 0; /* Size set later. */
check_dma(wn);
2005-04-21 16:53:53 +02:00
} else {
/* Not an ATA device; no translations, no special features. Don't
* touch it.
2005-04-21 16:53:53 +02:00
*/
wakeup_ticks = prev_wakeup;
w_testing = 0;
return(ERR);
2005-04-21 16:53:53 +02:00
}
/* Restore wakeup_ticks and unset testing mode. */
wakeup_ticks = prev_wakeup;
w_testing = 0;
2005-04-21 16:53:53 +02:00
/* Size of the whole drive */
wn->part[0].dv_size = (u64_t)size * SECTOR_SIZE;
2005-04-21 16:53:53 +02:00
/* Reset/calibrate (where necessary) */
if (w_specify() != OK && w_specify() != OK) {
return(ERR);
}
2005-04-21 16:53:53 +02:00
wn->state |= IDENTIFIED;
2005-04-21 16:53:53 +02:00
return(OK);
}
/*===========================================================================*
* w_name *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static char *w_name(void)
2005-04-21 16:53:53 +02:00
{
/* Return a name for the current device. */
static char name[] = "AT0-D0";
2005-04-21 16:53:53 +02:00
name[2] = '0' + w_instance;
name[5] = '0' + w_drive;
2005-04-21 16:53:53 +02:00
return name;
}
/*===========================================================================*
* w_io_test *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static int w_io_test(void)
{
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
int save_dev;
int save_timeout, save_errors, save_wakeup;
iovec_t iov;
static char *buf;
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
ssize_t r;
#define BUFSIZE CD_SECTOR_SIZE
STATICINIT(buf, BUFSIZE);
iov.iov_addr = (vir_bytes) buf;
iov.iov_size = BUFSIZE;
save_dev = w_device;
/* Reduce timeout values for this test transaction. */
save_timeout = timeout_usecs;
save_errors = max_errors;
save_wakeup = wakeup_ticks;
2005-09-11 19:09:11 +02:00
if (!w_standard_timeouts) {
timeout_usecs = 4000000;
wakeup_ticks = system_hz * 6;
max_errors = 3;
}
w_testing = 1;
/* Try I/O on the actual drive (not any (sub)partition). */
r = w_transfer(w_drive * DEV_PER_DRIVE, FALSE /*do_write*/, 0,
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
SELF, &iov, 1, BDEV_NOFLAGS);
/* Switch back. */
if (w_prepare(save_dev) == NULL)
panic("Couldn't switch back devices");
/* Restore parameters. */
timeout_usecs = save_timeout;
max_errors = save_errors;
wakeup_ticks = save_wakeup;
w_testing = 0;
/* Test if everything worked. */
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
if (r != BUFSIZE) {
return ERR;
}
/* Everything worked. */
return OK;
}
2005-04-21 16:53:53 +02:00
/*===========================================================================*
* w_specify *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static int w_specify(void)
2005-04-21 16:53:53 +02:00
{
/* Routine to initialize the drive after boot or when a reset is needed. */
struct wini *wn = w_wn;
struct command cmd;
2005-05-03 10:59:13 +02:00
if ((wn->state & DEAF) && w_reset() != OK) {
return(ERR);
}
2005-04-21 16:53:53 +02:00
if (!(wn->state & ATAPI)) {
/* Specify parameters: precompensation, number of heads and sectors. */
cmd.precomp = 0;
cmd.count = wn->sectors;
cmd.ldh = w_wn->ldhpref | (wn->heads - 1);
2005-04-21 16:53:53 +02:00
cmd.command = CMD_SPECIFY; /* Specify some parameters */
/* Output command block and see if controller accepts the parameters. */
if (com_simple(&cmd) != OK) return(ERR);
if (!(wn->state & SMART)) {
/* Calibrate an old disk. */
cmd.sector = 0;
cmd.cyl_lo = 0;
cmd.cyl_hi = 0;
cmd.ldh = w_wn->ldhpref;
cmd.command = CMD_RECALIBRATE;
if (com_simple(&cmd) != OK) return(ERR);
}
}
wn->state |= INITIALIZED;
return(OK);
}
2005-08-25 18:25:19 +02:00
/*===========================================================================*
* do_transfer *
*===========================================================================*/
static int do_transfer(const struct wini *wn, unsigned int count,
unsigned int sector, unsigned int do_write, int do_dma)
2005-08-25 18:25:19 +02:00
{
struct command cmd;
2006-02-24 13:55:47 +01:00
unsigned int sector_high;
unsigned secspcyl = wn->heads * wn->sectors;
2006-02-24 13:55:47 +01:00
int do_lba48;
sector_high= 0; /* For future extensions */
do_lba48= 0;
if (sector >= LBA48_CHECK_SIZE || sector_high != 0)
{
if (wn->lba48)
do_lba48= 1;
else if (sector > LBA_MAX_SIZE || sector_high != 0)
{
/* Strange sector count for LBA device */
return EIO;
}
}
2005-08-25 18:25:19 +02:00
cmd.precomp = 0;
2005-08-25 18:25:19 +02:00
cmd.count = count;
2006-03-03 16:37:51 +01:00
if (do_dma)
{
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
cmd.command = do_write ? CMD_WRITE_DMA : CMD_READ_DMA;
2006-03-03 16:37:51 +01:00
}
else
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
cmd.command = do_write ? CMD_WRITE : CMD_READ;
2006-02-24 13:55:47 +01:00
if (do_lba48) {
2006-03-03 16:37:51 +01:00
if (do_dma)
{
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
cmd.command = (do_write ?
2006-03-03 16:37:51 +01:00
CMD_WRITE_DMA_EXT : CMD_READ_DMA_EXT);
}
else
{
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
cmd.command = (do_write ?
2006-03-03 16:37:51 +01:00
CMD_WRITE_EXT : CMD_READ_EXT);
}
2006-02-24 13:55:47 +01:00
cmd.count_prev= (count >> 8);
cmd.sector = (sector >> 0) & 0xFF;
cmd.cyl_lo = (sector >> 8) & 0xFF;
cmd.cyl_hi = (sector >> 16) & 0xFF;
cmd.sector_prev= (sector >> 24) & 0xFF;
cmd.cyl_lo_prev= (sector_high) & 0xFF;
cmd.cyl_hi_prev= (sector_high >> 8) & 0xFF;
cmd.ldh = wn->ldhpref;
return com_out_ext(&cmd);
} else if (wn->ldhpref & LDH_LBA) {
2005-08-25 18:25:19 +02:00
cmd.sector = (sector >> 0) & 0xFF;
cmd.cyl_lo = (sector >> 8) & 0xFF;
cmd.cyl_hi = (sector >> 16) & 0xFF;
cmd.ldh = wn->ldhpref | ((sector >> 24) & 0xF);
} else {
int cylinder, head, sec;
cylinder = sector / secspcyl;
head = (sector % secspcyl) / wn->sectors;
sec = sector % wn->sectors;
2005-08-25 18:25:19 +02:00
cmd.sector = sec + 1;
cmd.cyl_lo = cylinder & BYTE;
cmd.cyl_hi = (cylinder >> 8) & BYTE;
cmd.ldh = wn->ldhpref | head;
}
return com_out(&cmd);
}
2012-03-25 20:25:53 +02:00
static void stop_dma(const struct wini *wn)
{
int r;
/* Stop bus master operation */
r= sys_outb(wn->base_dma + DMA_COMMAND, 0);
if (r != 0) panic("stop_dma: sys_outb failed: %d", r);
}
2012-03-25 20:25:53 +02:00
static void start_dma(const struct wini *wn, int do_write)
{
u32_t v;
int r;
/* Assume disk reads. Start DMA */
v= DMA_CMD_START;
if (!do_write)
{
/* Disk reads generate PCI write cycles. */
v |= DMA_CMD_WRITE;
}
r= sys_outb(wn->base_dma + DMA_COMMAND, v);
if (r != 0) panic("start_dma: sys_outb failed: %d", r);
}
2012-03-25 20:25:53 +02:00
static int error_dma(const struct wini *wn)
{
int r;
2012-03-05 00:11:41 +01:00
u32_t v;
#define DMAERR(msg) \
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
printf("at_wini%ld: bad DMA: %s. Disabling DMA for drive %d.\n", \
w_instance, msg, wn - wini); \
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
printf("at_wini%ld: workaround: set %s=1 in boot monitor.\n", \
w_instance, NO_DMA_VAR); \
return 1; \
r= sys_inb(wn->base_dma + DMA_STATUS, &v);
if (r != 0) panic("w_transfer: sys_inb failed: %d", r);
if (!wn->dma_intseen) {
/* DMA did not complete successfully */
if (v & DMA_ST_BM_ACTIVE) {
DMAERR("DMA did not complete");
} else if (v & DMA_ST_ERROR) {
DMAERR("DMA error");
} else {
DMAERR("DMA buffer too small");
}
} else if ((v & DMA_ST_BM_ACTIVE)) {
DMAERR("DMA buffer too large");
}
return 0;
}
2005-04-21 16:53:53 +02:00
/*===========================================================================*
* w_transfer *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static ssize_t w_transfer(
devminor_t minor, /* minor device to perform the transfer on */
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
int do_write, /* read or write? */
u64_t position, /* offset on device to read or write */
endpoint_t proc_nr, /* process doing the request */
iovec_t *iov, /* pointer to read or write request vector */
unsigned int nr_req, /* length of request vector */
int UNUSED(flags) /* transfer flags */
)
2005-04-21 16:53:53 +02:00
{
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
struct wini *wn;
2005-04-21 16:53:53 +02:00
iovec_t *iop, *iov_end = iov + nr_req;
int r, s, errors, do_dma;
2012-03-05 00:11:41 +01:00
unsigned long block;
u32_t w_status;
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
u64_t dv_size;
unsigned int n, nbytes;
2006-03-03 16:37:51 +01:00
unsigned dma_buf_offset;
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
ssize_t total = 0;
size_t addr_offset = 0;
2005-04-21 16:53:53 +02:00
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
if (w_prepare(minor) == NULL) return(ENXIO);
wn = w_wn;
dv_size = w_dv->dv_size;
2005-04-21 16:53:53 +02:00
if (w_wn->state & ATAPI) {
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
return atapi_transfer(do_write, position, proc_nr, iov, nr_req);
2005-04-21 16:53:53 +02:00
}
/* Check disk address. */
if ((unsigned)(position % SECTOR_SIZE) != 0) return(EINVAL);
2005-04-21 16:53:53 +02:00
errors = 0;
while (nr_req > 0) {
/* How many bytes to transfer? */
nbytes = 0;
for (iop = iov; iop < iov_end; iop++) nbytes += iop->iov_size;
if ((nbytes & SECTOR_MASK) != 0) return(EINVAL);
/* Which block on disk and how close to EOF? */
if (position >= dv_size) return(total); /* At EOF */
if (position + nbytes > dv_size)
nbytes = (unsigned)(dv_size - position);
block = (unsigned long)((w_dv->dv_base + position) / SECTOR_SIZE);
2005-04-21 16:53:53 +02:00
do_dma= wn->dma;
2006-03-03 16:37:51 +01:00
2005-04-21 16:53:53 +02:00
if (nbytes >= wn->max_count) {
/* The drive can't do more then max_count at once. */
nbytes = wn->max_count;
}
/* First check to see if a reinitialization is needed. */
if (!(wn->state & INITIALIZED) && w_specify() != OK) return(EIO);
if (do_dma) {
stop_dma(wn);
if (!setup_dma(&nbytes, proc_nr, iov, addr_offset, do_write)) {
do_dma = 0;
}
2006-03-03 16:37:51 +01:00
#if 0
printf("nbytes = %d\n", nbytes);
#endif
}
2005-04-21 16:53:53 +02:00
/* Tell the controller to transfer nbytes bytes. */
r = do_transfer(wn, (nbytes >> SECTOR_SHIFT), block, do_write, do_dma);
2005-04-21 16:53:53 +02:00
if (do_dma)
start_dma(wn, do_write);
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
if (do_write) {
2006-02-24 13:55:47 +01:00
/* The specs call for a 400 ns wait after issuing the command.
* Reading the alternate status register is the suggested
* way to implement this wait.
*/
if (sys_inb((wn->base_ctl+REG_CTL_ALTSTAT), &w_status) != OK)
panic("couldn't get status");
2006-02-24 13:55:47 +01:00
}
if (do_dma) {
2006-03-03 16:37:51 +01:00
/* Wait for the interrupt, check DMA status and optionally
* copy out.
*/
wn->dma_intseen = 0;
2006-03-03 16:37:51 +01:00
if ((r = at_intr_wait()) != OK)
{
/* Don't retry if sector marked bad or too many
* errors.
*/
if (r == ERR_BAD_SECTOR || ++errors == max_errors) {
w_command = CMD_IDLE;
return(EIO);
}
continue;
}
/* Wait for DMA_ST_INT to get set */
if (!wn->dma_intseen) {
if(w_waitfor_dma(DMA_ST_INT, DMA_ST_INT))
wn->dma_intseen = 1;
}
2009-04-27 13:53:11 +02:00
if (error_dma(wn)) {
wn->dma = 0;
continue;
2009-04-27 13:53:11 +02:00
}
2006-03-03 16:37:51 +01:00
stop_dma(wn);
2006-03-03 16:37:51 +01:00
dma_buf_offset= 0;
while (r == OK && nbytes > 0)
{
n= iov->iov_size;
if (n > nbytes)
n= nbytes;
/* Book the bytes successfully transferred. */
nbytes -= n;
position= position + n;
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
total += n;
addr_offset += n;
if ((iov->iov_size -= n) == 0) {
iov++; nr_req--; addr_offset = 0;
}
2006-03-03 16:37:51 +01:00
dma_buf_offset += n;
}
}
2005-04-21 16:53:53 +02:00
while (r == OK && nbytes > 0) {
/* For each sector, wait for an interrupt and fetch the data
* (read), or supply data to the controller and wait for an
* interrupt (write).
*/
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
if (!do_write) {
2005-04-21 16:53:53 +02:00
/* First an interrupt, then data. */
if ((r = at_intr_wait()) != OK) {
/* An error, send data to the bit bucket. */
if (w_wn->w_status & STATUS_DRQ) {
2006-02-24 13:55:47 +01:00
if ((s=sys_insw(wn->base_cmd+REG_DATA,
SELF, tmp_buf,
SECTOR_SIZE)) != OK) {
panic("Call to sys_insw() failed: %d", s);
2006-02-24 13:55:47 +01:00
}
2005-04-21 16:53:53 +02:00
}
break;
}
}
2006-02-24 13:55:47 +01:00
/* Wait for busy to clear. */
if (!w_waitfor(STATUS_BSY, 0)) { r = ERR; break; }
2005-04-21 16:53:53 +02:00
/* Wait for data transfer requested. */
if (!w_waitfor(STATUS_DRQ, STATUS_DRQ)) { r = ERR; break; }
/* Copy bytes to or from the device's buffer. */
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
if (!do_write) {
if(proc_nr != SELF) {
s=sys_safe_insw(wn->base_cmd + REG_DATA, proc_nr,
(void *) (iov->iov_addr), addr_offset,
SECTOR_SIZE);
} else {
s=sys_insw(wn->base_cmd + REG_DATA, proc_nr,
(void *) (iov->iov_addr + addr_offset),
SECTOR_SIZE);
}
if(s != OK) {
panic("Call to sys_insw() failed: %d", s);
}
2005-04-21 16:53:53 +02:00
} else {
if(proc_nr != SELF) {
s=sys_safe_outsw(wn->base_cmd + REG_DATA, proc_nr,
(void *) (iov->iov_addr), addr_offset,
SECTOR_SIZE);
} else {
s=sys_outsw(wn->base_cmd + REG_DATA, proc_nr,
(void *) (iov->iov_addr + addr_offset),
SECTOR_SIZE);
}
if(s != OK) {
panic("Call to sys_outsw() failed: %d", s);
}
2005-04-21 16:53:53 +02:00
/* Data sent, wait for an interrupt. */
if ((r = at_intr_wait()) != OK) break;
2005-04-21 16:53:53 +02:00
}
/* Book the bytes successfully transferred. */
nbytes -= SECTOR_SIZE;
position = position + SECTOR_SIZE;
addr_offset += SECTOR_SIZE;
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
total += SECTOR_SIZE;
if ((iov->iov_size -= SECTOR_SIZE) == 0) {
iov++;
nr_req--;
addr_offset = 0;
}
2005-04-21 16:53:53 +02:00
}
/* Any errors? */
if (r != OK) {
/* Don't retry if sector marked bad or too many errors. */
if (r == ERR_BAD_SECTOR || ++errors == max_errors) {
2005-04-21 16:53:53 +02:00
w_command = CMD_IDLE;
return(EIO);
}
}
}
w_command = CMD_IDLE;
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
return(total);
2005-04-21 16:53:53 +02:00
}
2005-09-11 19:09:11 +02:00
/*===========================================================================*
* com_out *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static int com_out(cmd)
2005-04-21 16:53:53 +02:00
struct command *cmd; /* Command block */
{
/* Output the command block to the winchester controller and return status */
struct wini *wn = w_wn;
2005-08-25 18:25:19 +02:00
unsigned base_cmd = wn->base_cmd;
unsigned base_ctl = wn->base_ctl;
2005-04-21 16:53:53 +02:00
pvb_pair_t outbyte[7]; /* vector for sys_voutb() */
int s; /* status for sys_(v)outb() */
2005-09-11 19:09:11 +02:00
if (w_wn->state & IGNORING) return ERR;
2005-04-21 16:53:53 +02:00
if (!w_waitfor(STATUS_BSY, 0)) {
printf("%s: controller not ready\n", w_name());
return(ERR);
}
/* Select drive. */
2005-08-25 18:25:19 +02:00
if ((s=sys_outb(base_cmd + REG_LDH, cmd->ldh)) != OK)
panic("Couldn't write register to select drive: %d", s);
2005-04-21 16:53:53 +02:00
if (!w_waitfor(STATUS_BSY, 0)) {
2005-05-03 10:59:13 +02:00
printf("%s: com_out: drive not ready\n", w_name());
2005-04-21 16:53:53 +02:00
return(ERR);
}
/* Schedule a wakeup call, some controllers are flaky. This is done with a
* synchronous alarm. If a timeout occurs a notify from CLOCK is sent, so that
* w_intr_wait() can call w_timeout() in case the controller was not able to
* execute the command. Leftover timeouts are simply ignored by the main loop.
2005-04-21 16:53:53 +02:00
*/
sys_setalarm(wakeup_ticks, 0);
2005-04-21 16:53:53 +02:00
wn->w_status = STATUS_ADMBSY;
2005-04-21 16:53:53 +02:00
w_command = cmd->command;
pv_set(outbyte[0], base_ctl + REG_CTL, wn->heads >= 8 ? CTL_EIGHTHEADS : 0);
2005-08-25 18:25:19 +02:00
pv_set(outbyte[1], base_cmd + REG_PRECOMP, cmd->precomp);
pv_set(outbyte[2], base_cmd + REG_COUNT, cmd->count);
pv_set(outbyte[3], base_cmd + REG_SECTOR, cmd->sector);
pv_set(outbyte[4], base_cmd + REG_CYL_LO, cmd->cyl_lo);
pv_set(outbyte[5], base_cmd + REG_CYL_HI, cmd->cyl_hi);
pv_set(outbyte[6], base_cmd + REG_COMMAND, cmd->command);
2005-04-21 16:53:53 +02:00
if ((s=sys_voutb(outbyte,7)) != OK)
panic("Couldn't write registers with sys_voutb(): %d", s);
2005-04-21 16:53:53 +02:00
return(OK);
}
2006-02-24 13:55:47 +01:00
/*===========================================================================*
* com_out_ext *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static int com_out_ext(cmd)
2006-02-24 13:55:47 +01:00
struct command *cmd; /* Command block */
{
/* Output the command block to the winchester controller and return status */
struct wini *wn = w_wn;
unsigned base_cmd = wn->base_cmd;
unsigned base_ctl = wn->base_ctl;
pvb_pair_t outbyte[11]; /* vector for sys_voutb() */
int s; /* status for sys_(v)outb() */
if (w_wn->state & IGNORING) return ERR;
if (!w_waitfor(STATUS_BSY, 0)) {
printf("%s: controller not ready\n", w_name());
return(ERR);
}
/* Select drive. */
if ((s=sys_outb(base_cmd + REG_LDH, cmd->ldh)) != OK)
panic("Couldn't write register to select drive: %d", s);
2006-02-24 13:55:47 +01:00
if (!w_waitfor(STATUS_BSY, 0)) {
printf("%s: com_out: drive not ready\n", w_name());
return(ERR);
}
/* Schedule a wakeup call, some controllers are flaky. This is done with a
* synchronous alarm. If a timeout occurs a notify from CLOCK is sent, so that
* w_intr_wait() can call w_timeout() in case the controller was not able to
* execute the command. Leftover timeouts are simply ignored by the main loop.
2006-02-24 13:55:47 +01:00
*/
sys_setalarm(wakeup_ticks, 0);
wn->w_status = STATUS_ADMBSY;
w_command = cmd->command;
pv_set(outbyte[0], base_ctl + REG_CTL, 0);
pv_set(outbyte[1], base_cmd + REG_COUNT, cmd->count_prev);
pv_set(outbyte[2], base_cmd + REG_SECTOR, cmd->sector_prev);
pv_set(outbyte[3], base_cmd + REG_CYL_LO, cmd->cyl_lo_prev);
pv_set(outbyte[4], base_cmd + REG_CYL_HI, cmd->cyl_hi_prev);
pv_set(outbyte[5], base_cmd + REG_COUNT, cmd->count);
pv_set(outbyte[6], base_cmd + REG_SECTOR, cmd->sector);
pv_set(outbyte[7], base_cmd + REG_CYL_LO, cmd->cyl_lo);
pv_set(outbyte[8], base_cmd + REG_CYL_HI, cmd->cyl_hi);
2007-02-23 21:58:10 +01:00
pv_set(outbyte[9], base_cmd + REG_COMMAND, cmd->command);
if ((s=sys_voutb(outbyte, 10)) != OK)
panic("Couldn't write registers with sys_voutb(): %d", s);
2006-02-24 13:55:47 +01:00
return(OK);
}
2006-03-03 16:37:51 +01:00
/*===========================================================================*
* setup_dma *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static int setup_dma(
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
unsigned *sizep,
endpoint_t proc_nr,
iovec_t *iov,
size_t addr_offset,
int UNUSED(do_write)
)
2006-03-03 16:37:51 +01:00
{
phys_bytes user_phys;
2006-03-03 16:37:51 +01:00
unsigned n, offset, size;
int i, j, r;
2012-03-05 00:11:41 +01:00
u32_t v;
2006-03-03 16:37:51 +01:00
struct wini *wn = w_wn;
/* First try direct scatter/gather to the supplied buffers */
size= *sizep;
i= 0; /* iov index */
j= 0; /* prdt index */
offset= 0; /* Offset in current iov */
2012-07-27 16:49:55 +02:00
#if VERBOSE_DMA
printf("at_wini: setup_dma: proc_nr %d\n", proc_nr);
#endif
2006-03-03 16:37:51 +01:00
while (size > 0)
{
2012-07-27 16:49:55 +02:00
#if VERBOSE_DMA
printf(
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
"at_wini: setup_dma: iov[%d]: addr 0x%lx, size %ld offset %d, size %d\n",
2006-03-03 16:37:51 +01:00
i, iov[i].iov_addr, iov[i].iov_size, offset, size);
2012-07-27 16:49:55 +02:00
#endif
2006-03-03 16:37:51 +01:00
n= iov[i].iov_size-offset;
if (n > size)
n= size;
if (n == 0 || (n & 1))
panic("bad size in iov: 0x%lx", iov[i].iov_size);
if(proc_nr != SELF) {
r= sys_umap(proc_nr, VM_GRANT, iov[i].iov_addr, n,
&user_phys);
if (r != 0)
panic("can't map user buffer (VM_GRANT): %d", r);
user_phys += offset + addr_offset;
} else {
r= sys_umap(proc_nr, VM_D,
iov[i].iov_addr+offset+addr_offset, n,
&user_phys);
if (r != 0)
panic("can't map user buffer (VM_D): %d", r);
}
2006-03-03 16:37:51 +01:00
if (user_phys & 1)
{
/* Buffer is not aligned */
printf("setup_dma: user buffer is not aligned\n");
return 0;
2006-03-03 16:37:51 +01:00
}
/* vector is not allowed to cross a 64K boundary */
if (user_phys/0x10000 != (user_phys+n-1)/0x10000)
n= ((user_phys/0x10000)+1)*0x10000 - user_phys;
/* vector is not allowed to be bigger than 64K, but we get that
* for free.
*/
if (j >= N_PRDTE)
{
/* Too many entries */
printf("setup_dma: user buffer has too many entries\n");
return 0;
2006-03-03 16:37:51 +01:00
}
prdt[j].prdte_base= user_phys;
prdt[j].prdte_count= n;
prdt[j].prdte_reserved= 0;
prdt[j].prdte_flags= 0;
j++;
offset += n;
if (offset >= iov[i].iov_size)
{
i++;
offset= 0;
addr_offset= 0;
2006-03-03 16:37:51 +01:00
}
size -= n;
}
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
if (j <= 0 || j > N_PRDTE)
panic("bad prdt index: %d", j);
prdt[j-1].prdte_flags |= PRDTE_FL_EOT;
2006-03-03 16:37:51 +01:00
2012-07-27 16:49:55 +02:00
#if VERBOSE_DMA
printf("dma not bad\n");
for (i= 0; i<j; i++) {
printf("prdt[%d]: base 0x%lx, size %d, flags 0x%x\n",
i, prdt[i].prdte_base, prdt[i].prdte_count,
prdt[i].prdte_flags);
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
}
2012-07-27 16:49:55 +02:00
#endif
2006-03-03 16:37:51 +01:00
/* Verify that the bus master is not active */
r= sys_inb(wn->base_dma + DMA_STATUS, &v);
if (r != 0) panic("setup_dma: sys_inb failed: %d", r);
2006-03-03 16:37:51 +01:00
if (v & DMA_ST_BM_ACTIVE)
panic("Bus master IDE active");
2006-03-03 16:37:51 +01:00
if (prdt_phys & 3)
panic("prdt not aligned: 0x%lx", prdt_phys);
2006-03-03 16:37:51 +01:00
r= sys_outl(wn->base_dma + DMA_PRDTP, prdt_phys);
if (r != 0) panic("setup_dma: sys_outl failed: %d", r);
2006-03-03 16:37:51 +01:00
/* Clear interrupt and error flags */
r= sys_outb(wn->base_dma + DMA_STATUS, DMA_ST_INT | DMA_ST_ERROR);
if (r != 0) panic("setup_dma: sys_outb failed: %d", r);
return 1;
2006-03-03 16:37:51 +01:00
}
2005-04-21 16:53:53 +02:00
/*===========================================================================*
* w_need_reset *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static void w_need_reset(void)
2005-04-21 16:53:53 +02:00
{
/* The controller needs to be reset. */
struct wini *wn;
for (wn = wini; wn < &wini[MAX_DRIVES]; wn++) {
2005-08-25 18:25:19 +02:00
if (wn->base_cmd == w_wn->base_cmd) {
wn->state |= DEAF;
wn->state &= ~INITIALIZED;
}
2005-04-21 16:53:53 +02:00
}
}
2005-09-11 19:09:11 +02:00
/*===========================================================================*
* w_do_close *
*===========================================================================*/
static int w_do_close(devminor_t minor)
2005-04-21 16:53:53 +02:00
{
/* Device close: Release a device. */
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
if (w_prepare(minor) == NULL)
return(ENXIO);
2005-04-21 16:53:53 +02:00
w_wn->open_ct--;
if (w_wn->open_ct == 0 && (w_wn->state & ATAPI)) atapi_close();
return(OK);
}
2005-09-11 19:09:11 +02:00
/*===========================================================================*
* com_simple *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static int com_simple(cmd)
2005-04-21 16:53:53 +02:00
struct command *cmd; /* Command block */
{
/* A simple controller command, only one interrupt and no data-out phase. */
int r;
2005-09-11 19:09:11 +02:00
if (w_wn->state & IGNORING) return ERR;
2005-04-21 16:53:53 +02:00
if ((r = com_out(cmd)) == OK) r = at_intr_wait();
w_command = CMD_IDLE;
return(r);
}
/*===========================================================================*
* w_timeout *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static void w_timeout(void)
2005-04-21 16:53:53 +02:00
{
struct wini *wn = w_wn;
switch (w_command) {
case CMD_IDLE:
break; /* fine */
case CMD_READ:
2006-02-24 13:55:47 +01:00
case CMD_READ_EXT:
2005-04-21 16:53:53 +02:00
case CMD_WRITE:
2006-02-24 13:55:47 +01:00
case CMD_WRITE_EXT:
2005-04-21 16:53:53 +02:00
/* Impossible, but not on PC's: The controller does not respond. */
/* Limiting multisector I/O seems to help. */
if (wn->max_count > 8 * SECTOR_SIZE) {
wn->max_count = 8 * SECTOR_SIZE;
} else {
wn->max_count = SECTOR_SIZE;
}
/*FALL THROUGH*/
default:
/* Some other command. */
2005-09-11 19:09:11 +02:00
if (w_testing) wn->state |= IGNORING; /* Kick out this drive. */
2006-02-24 13:55:47 +01:00
else if (!w_silent) printf("%s: timeout on command 0x%02x\n",
w_name(), w_command);
2005-04-21 16:53:53 +02:00
w_need_reset();
wn->w_status = 0;
2005-04-21 16:53:53 +02:00
}
}
/*===========================================================================*
* w_reset *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static int w_reset(void)
2005-04-21 16:53:53 +02:00
{
/* Issue a reset to the controller. This is done after any catastrophe,
* like the controller refusing to respond.
*/
int s;
struct wini *wn = w_wn;
/* Don't bother if this drive is forgotten. */
2005-09-11 19:09:11 +02:00
if (w_wn->state & IGNORING) return ERR;
2005-04-21 16:53:53 +02:00
/* Wait for any internal drive recovery. */
tickdelay(RECOVERY_TICKS);
2005-04-21 16:53:53 +02:00
/* Strobe reset bit */
2005-08-25 18:25:19 +02:00
if ((s=sys_outb(wn->base_ctl + REG_CTL, CTL_RESET)) != OK)
panic("Couldn't strobe reset bit: %d", s);
tickdelay(DELAY_TICKS);
2005-08-25 18:25:19 +02:00
if ((s=sys_outb(wn->base_ctl + REG_CTL, 0)) != OK)
panic("Couldn't strobe reset bit: %d", s);
tickdelay(DELAY_TICKS);
2005-04-21 16:53:53 +02:00
/* Wait for controller ready */
if (!w_waitfor(STATUS_BSY, 0)) {
printf("%s: reset failed, drive busy\n", w_name());
return(ERR);
}
/* The error register should be checked now, but some drives mess it up. */
for (wn = wini; wn < &wini[MAX_DRIVES]; wn++) {
2005-08-25 18:25:19 +02:00
if (wn->base_cmd == w_wn->base_cmd) {
wn->state &= ~DEAF;
if (w_wn->native) {
/* Make sure irq is actually enabled.. */
sys_irqenable(&w_wn->irq_hook_id);
}
}
2005-04-21 16:53:53 +02:00
}
2005-04-21 16:53:53 +02:00
return(OK);
}
2005-09-11 19:09:11 +02:00
/*===========================================================================*
* w_intr_wait *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static void w_intr_wait(void)
2005-04-21 16:53:53 +02:00
{
/* Wait for a task completion interrupt. */
2006-02-24 13:55:47 +01:00
int r;
2012-03-05 00:11:41 +01:00
u32_t w_status;
2005-04-21 16:53:53 +02:00
message m;
Driver refactory for live update and crash recovery. SYSLIB CHANGES: - DS calls to publish / retrieve labels consider endpoints instead of u32_t. VFS CHANGES: - mapdriver() only adds an entry in the dmap table in VFS. - dev_up() is only executed upon reception of a driver up event. INET CHANGES: - INET no longer searches for existing drivers instances at startup. - A newtwork driver is (re)initialized upon reception of a driver up event. - Networking startup is now race-free by design. No need to waste 5 seconds at startup any more. DRIVER CHANGES: - Every driver publishes driver up events when starting for the first time or in case of restart when recovery actions must be taken in the upper layers. - Driver up events are published by drivers through DS. - For regular drivers, VFS is normally the only subscriber, but not necessarily. For instance, when the filter driver is in use, it must subscribe to driver up events to initiate recovery. - For network drivers, inet is the only subscriber for now. - Every VFS driver is statically linked with libdriver, every network driver is statically linked with libnetdriver. DRIVER LIBRARIES CHANGES: - Libdriver is extended to provide generic receive() and ds_publish() interfaces for VFS drivers. - driver_receive() is a wrapper for sef_receive() also used in driver_task() to discard spurious messages that were meant to be delivered to a previous version of the driver. - driver_receive_mq() is the same as driver_receive() but integrates support for queued messages. - driver_announce() publishes a driver up event for VFS drivers and marks the driver as initialized and expecting a DEV_OPEN message. - Libnetdriver is introduced to provide similar receive() and ds_publish() interfaces for network drivers (netdriver_announce() and netdriver_receive()). - Network drivers all support live update with no state transfer now. KERNEL CHANGES: - Added kernel call statectl for state management. Used by driver_announce() to unblock eventual callers sendrecing to the driver.
2010-04-08 15:41:35 +02:00
int ipc_status;
2005-04-21 16:53:53 +02:00
if (w_wn->state & IDENTIFIED) {
/* Wait for an interrupt that sets w_status to "not busy".
* (w_timeout() also clears w_status.)
*/
while (w_wn->w_status & (STATUS_ADMBSY|STATUS_BSY)) {
if ((r=driver_receive(ANY, &m, &ipc_status)) != OK)
panic("driver_receive failed: %d", r);
Driver refactory for live update and crash recovery. SYSLIB CHANGES: - DS calls to publish / retrieve labels consider endpoints instead of u32_t. VFS CHANGES: - mapdriver() only adds an entry in the dmap table in VFS. - dev_up() is only executed upon reception of a driver up event. INET CHANGES: - INET no longer searches for existing drivers instances at startup. - A newtwork driver is (re)initialized upon reception of a driver up event. - Networking startup is now race-free by design. No need to waste 5 seconds at startup any more. DRIVER CHANGES: - Every driver publishes driver up events when starting for the first time or in case of restart when recovery actions must be taken in the upper layers. - Driver up events are published by drivers through DS. - For regular drivers, VFS is normally the only subscriber, but not necessarily. For instance, when the filter driver is in use, it must subscribe to driver up events to initiate recovery. - For network drivers, inet is the only subscriber for now. - Every VFS driver is statically linked with libdriver, every network driver is statically linked with libnetdriver. DRIVER LIBRARIES CHANGES: - Libdriver is extended to provide generic receive() and ds_publish() interfaces for VFS drivers. - driver_receive() is a wrapper for sef_receive() also used in driver_task() to discard spurious messages that were meant to be delivered to a previous version of the driver. - driver_receive_mq() is the same as driver_receive() but integrates support for queued messages. - driver_announce() publishes a driver up event for VFS drivers and marks the driver as initialized and expecting a DEV_OPEN message. - Libnetdriver is introduced to provide similar receive() and ds_publish() interfaces for network drivers (netdriver_announce() and netdriver_receive()). - Network drivers all support live update with no state transfer now. KERNEL CHANGES: - Added kernel call statectl for state management. Used by driver_announce() to unblock eventual callers sendrecing to the driver.
2010-04-08 15:41:35 +02:00
if (is_ipc_notify(ipc_status)) {
switch (_ENDPOINT_P(m.m_source)) {
case CLOCK:
/* Timeout. */
w_timeout(); /* a.o. set w_status */
break;
case HARDWARE:
/* Interrupt. */
r= sys_inb(w_wn->base_cmd +
REG_STATUS, &w_status);
if (r != 0)
panic("sys_inb failed: %d", r);
w_wn->w_status= w_status;
w_hw_int(m.m_notify.interrupts);
break;
default:
/*
* unhandled message. queue it and
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
* handle it in the blockdriver loop.
*/
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
blockdriver_mq_queue(&m, ipc_status);
}
}
else {
/*
* unhandled message. queue it and handle it in the
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
* blockdriver loop.
*/
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
blockdriver_mq_queue(&m, ipc_status);
2006-03-07 16:45:14 +01:00
}
2005-04-21 16:53:53 +02:00
}
} else {
/* Device not yet identified; use polling. */
2005-04-21 16:53:53 +02:00
(void) w_waitfor(STATUS_BSY, 0);
}
}
2005-09-11 19:09:11 +02:00
/*===========================================================================*
* at_intr_wait *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static int at_intr_wait(void)
2005-04-21 16:53:53 +02:00
{
/* Wait for an interrupt, study the status bits and return error/success. */
2006-02-24 13:55:47 +01:00
int r, s;
2012-03-05 00:11:41 +01:00
u32_t inbval;
2005-04-21 16:53:53 +02:00
w_intr_wait();
if ((w_wn->w_status & (STATUS_BSY | STATUS_WF | STATUS_ERR)) == 0) {
2005-04-21 16:53:53 +02:00
r = OK;
} else {
2005-08-25 18:25:19 +02:00
if ((s=sys_inb(w_wn->base_cmd + REG_ERROR, &inbval)) != OK)
panic("Couldn't read register: %d", s);
if ((w_wn->w_status & STATUS_ERR) && (inbval & ERROR_BB)) {
2005-04-21 16:53:53 +02:00
r = ERR_BAD_SECTOR; /* sector marked bad, retries won't help */
} else {
r = ERR; /* any other error */
}
}
w_wn->w_status |= STATUS_ADMBSY; /* assume still busy with I/O */
2005-04-21 16:53:53 +02:00
return(r);
}
2005-09-11 19:09:11 +02:00
/*===========================================================================*
* w_waitfor *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static int w_waitfor(mask, value)
2005-04-21 16:53:53 +02:00
int mask; /* status mask */
int value; /* required status */
{
/* Wait until controller is in the required state. Return zero on timeout.
*/
2012-03-05 00:11:41 +01:00
u32_t w_status;
spin_t spin;
2005-04-21 16:53:53 +02:00
int s;
2006-02-24 13:55:47 +01:00
SPIN_FOR(&spin, timeout_usecs) {
2006-02-24 13:55:47 +01:00
if ((s=sys_inb(w_wn->base_cmd + REG_STATUS, &w_status)) != OK)
panic("Couldn't read register: %d", s);
2006-02-24 13:55:47 +01:00
w_wn->w_status= w_status;
if ((w_wn->w_status & mask) == value) {
2005-04-21 16:53:53 +02:00
return 1;
}
}
2005-04-21 16:53:53 +02:00
w_need_reset(); /* controller gone deaf */
return(0);
}
2006-03-03 16:37:51 +01:00
/*===========================================================================*
* w_waitfor_dma *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static int w_waitfor_dma(mask, value)
unsigned int mask; /* status mask */
unsigned value; /* required status */
2006-03-03 16:37:51 +01:00
{
/* Wait until controller is in the required state. Return zero on timeout.
*/
2012-03-05 00:11:41 +01:00
u32_t w_status;
spin_t spin;
2006-03-03 16:37:51 +01:00
int s;
SPIN_FOR(&spin, timeout_usecs) {
2006-03-03 16:37:51 +01:00
if ((s=sys_inb(w_wn->base_dma + DMA_STATUS, &w_status)) != OK)
panic("Couldn't read register: %d", s);
2006-03-03 16:37:51 +01:00
if ((w_status & mask) == value) {
return 1;
}
}
2006-03-03 16:37:51 +01:00
return(0);
}
2005-09-11 19:09:11 +02:00
/*===========================================================================*
* w_geometry *
*===========================================================================*/
static void w_geometry(devminor_t minor, struct part_geom *entry)
2005-04-21 16:53:53 +02:00
{
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
struct wini *wn;
if (w_prepare(minor) == NULL) return;
wn = w_wn;
2005-04-21 16:53:53 +02:00
if (wn->state & ATAPI) { /* Make up some numbers. */
entry->cylinders = (unsigned long)(wn->part[0].dv_size / SECTOR_SIZE) / (64*32);
2005-04-21 16:53:53 +02:00
entry->heads = 64;
entry->sectors = 32;
} else { /* Return logical geometry. */
entry->cylinders = wn->cylinders;
entry->heads = wn->heads;
entry->sectors = wn->sectors;
while (entry->cylinders > 1024) {
entry->heads *= 2;
entry->cylinders /= 2;
}
2005-04-21 16:53:53 +02:00
}
}
/*===========================================================================*
* atapi_open *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static int atapi_open(void)
2005-04-21 16:53:53 +02:00
{
/* Should load and lock the device and obtain its size. For now just set the
* size of the device to something big. What is really needed is a generic
* SCSI layer that does all this stuff for ATAPI and SCSI devices (kjb). (XXX)
*/
w_wn->part[0].dv_size = (u64_t)(800L*1024) * 1024;
2005-04-21 16:53:53 +02:00
return(OK);
}
/*===========================================================================*
* atapi_close *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static void atapi_close(void)
2005-04-21 16:53:53 +02:00
{
/* Should unlock the device. For now do nothing. (XXX) */
}
2012-03-25 20:25:53 +02:00
static void sense_request(void)
{
int r, i;
static u8_t sense[100], packet[ATAPI_PACKETSIZE];
packet[0] = SCSI_SENSE;
packet[1] = 0;
packet[2] = 0;
packet[3] = 0;
packet[4] = SENSE_PACKETSIZE;
packet[5] = 0;
packet[7] = 0;
packet[8] = 0;
packet[9] = 0;
packet[10] = 0;
packet[11] = 0;
for(i = 0; i < SENSE_PACKETSIZE; i++) sense[i] = 0xff;
r = atapi_sendpacket(packet, SENSE_PACKETSIZE, 0);
if (r != OK) { printf("request sense command failed\n"); return; }
if (atapi_intr_wait(0, 0) <= 0) { printf("WARNING: request response failed\n"); }
if (sys_insw(w_wn->base_cmd + REG_DATA, SELF, (void *) sense, SENSE_PACKETSIZE) != OK)
printf("WARNING: sense reading failed\n");
printf("sense data:");
for(i = 0; i < SENSE_PACKETSIZE; i++) printf(" %02x", sense[i]);
printf("\n");
}
2005-04-21 16:53:53 +02:00
/*===========================================================================*
* atapi_transfer *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static int atapi_transfer(
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
int do_write, /* read or write? */
u64_t position, /* offset on device to read or write */
endpoint_t proc_nr, /* process doing the request */
iovec_t *iov, /* pointer to read or write request vector */
unsigned int nr_req /* length of request vector */
)
2005-04-21 16:53:53 +02:00
{
struct wini *wn = w_wn;
iovec_t *iop, *iov_end = iov + nr_req;
int r, s, errors, fresh;
u64_t pos;
unsigned long block;
u64_t dv_size = w_dv->dv_size;
unsigned nbytes, nblocks, before, chunk;
static u8_t packet[ATAPI_PACKETSIZE];
size_t addr_offset = 0;
int dmabytes = 0, piobytes = 0;
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
ssize_t total = 0;
if (do_write) return(EINVAL);
2005-04-21 16:53:53 +02:00
errors = fresh = 0;
while (nr_req > 0 && !fresh) {
int do_dma = wn->dma && w_atapi_dma;
2005-04-21 16:53:53 +02:00
/* The Minix block size is smaller than the CD block size, so we
* may have to read extra before or after the good data.
*/
pos = w_dv->dv_base + position;
block = (unsigned long)(pos / CD_SECTOR_SIZE);
before = (unsigned)(pos % CD_SECTOR_SIZE);
2005-04-21 16:53:53 +02:00
if (before)
do_dma = 0;
2005-04-21 16:53:53 +02:00
/* How many bytes to transfer? */
nbytes = 0;
2005-04-21 16:53:53 +02:00
for (iop = iov; iop < iov_end; iop++) {
nbytes += iop->iov_size;
if (iop->iov_size % CD_SECTOR_SIZE)
do_dma = 0;
2005-04-21 16:53:53 +02:00
}
/* Data comes in as words, so we have to enforce even byte counts. */
if ((before | nbytes) & 1) return(EINVAL);
/* Which block on disk and how close to EOF? */
if (position >= dv_size) return(total); /* At EOF */
if (position + nbytes > dv_size)
nbytes = (unsigned)(dv_size - position);
2005-04-21 16:53:53 +02:00
nblocks = (before + nbytes + CD_SECTOR_SIZE - 1) / CD_SECTOR_SIZE;
/* First check to see if a reinitialization is needed. */
if (!(wn->state & INITIALIZED) && w_specify() != OK) return(EIO);
/* Build an ATAPI command packet. */
packet[0] = SCSI_READ10;
packet[1] = 0;
packet[2] = (block >> 24) & 0xFF;
packet[3] = (block >> 16) & 0xFF;
packet[4] = (block >> 8) & 0xFF;
packet[5] = (block >> 0) & 0xFF;
packet[6] = 0;
2005-04-21 16:53:53 +02:00
packet[7] = (nblocks >> 8) & 0xFF;
packet[8] = (nblocks >> 0) & 0xFF;
packet[9] = 0;
packet[10] = 0;
packet[11] = 0;
if(do_dma) {
stop_dma(wn);
if (!setup_dma(&nbytes, proc_nr, iov, addr_offset, 0)) {
do_dma = 0;
} else if(nbytes != nblocks * CD_SECTOR_SIZE) {
stop_dma(wn);
do_dma = 0;
}
}
2005-04-21 16:53:53 +02:00
/* Tell the controller to execute the packet command. */
r = atapi_sendpacket(packet, nblocks * CD_SECTOR_SIZE, do_dma);
2005-04-21 16:53:53 +02:00
if (r != OK) goto err;
if(do_dma) {
wn->dma_intseen = 0;
start_dma(wn, 0);
w_intr_wait();
if(!wn->dma_intseen) {
if(w_waitfor_dma(DMA_ST_INT, DMA_ST_INT)) {
wn->dma_intseen = 1;
}
}
if(error_dma(wn)) {
printf("Disabling DMA (ATAPI)\n");
wn->dma = 0;
} else {
dmabytes += nbytes;
while (nbytes > 0) {
chunk = nbytes;
if (chunk > iov->iov_size)
chunk = iov->iov_size;
position = position + chunk;
nbytes -= chunk;
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
total += chunk;
if ((iov->iov_size -= chunk) == 0) {
iov++;
nr_req--;
}
}
}
continue;
}
2005-04-21 16:53:53 +02:00
/* Read chunks of data. */
while ((r = atapi_intr_wait(do_dma, nblocks * CD_SECTOR_SIZE)) > 0) {
size_t count;
2005-04-21 16:53:53 +02:00
count = r;
while (before > 0 && count > 0) { /* Discard before. */
chunk = before;
if (chunk > count) chunk = count;
if (chunk > DMA_BUF_SIZE) chunk = DMA_BUF_SIZE;
if ((s=sys_insw(wn->base_cmd + REG_DATA,
SELF, tmp_buf, chunk)) != OK)
panic("Call to sys_insw() failed: %d", s);
2005-04-21 16:53:53 +02:00
before -= chunk;
count -= chunk;
}
while (nbytes > 0 && count > 0) { /* Requested data. */
chunk = nbytes;
if (chunk > count) chunk = count;
if (chunk > iov->iov_size) chunk = iov->iov_size;
if(proc_nr != SELF) {
s=sys_safe_insw(wn->base_cmd + REG_DATA,
proc_nr, (void *) iov->iov_addr,
addr_offset, chunk);
} else {
s=sys_insw(wn->base_cmd + REG_DATA, proc_nr,
(void *) (iov->iov_addr + addr_offset),
chunk);
}
if (s != OK)
panic("Call to sys_insw() failed: %d", s);
position = position + chunk;
2005-04-21 16:53:53 +02:00
nbytes -= chunk;
count -= chunk;
addr_offset += chunk;
piobytes += chunk;
2005-04-21 16:53:53 +02:00
fresh = 0;
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
total += chunk;
2005-04-21 16:53:53 +02:00
if ((iov->iov_size -= chunk) == 0) {
iov++;
nr_req--;
fresh = 1; /* new element is optional */
addr_offset = 0;
2005-04-21 16:53:53 +02:00
}
2005-04-21 16:53:53 +02:00
}
while (count > 0) { /* Excess data. */
chunk = count;
if (chunk > DMA_BUF_SIZE) chunk = DMA_BUF_SIZE;
if ((s=sys_insw(wn->base_cmd + REG_DATA,
SELF, tmp_buf, chunk)) != OK)
panic("Call to sys_insw() failed: %d", s);
2005-04-21 16:53:53 +02:00
count -= chunk;
}
}
if (r < 0) {
err: /* Don't retry if too many errors. */
2005-09-11 19:09:11 +02:00
if (atapi_debug) sense_request();
if (++errors == max_errors) {
2005-04-21 16:53:53 +02:00
w_command = CMD_IDLE;
2005-09-11 19:09:11 +02:00
if (atapi_debug) printf("giving up (%d)\n", errors);
2005-04-21 16:53:53 +02:00
return(EIO);
}
2005-09-11 19:09:11 +02:00
if (atapi_debug) printf("retry (%d)\n", errors);
2005-04-21 16:53:53 +02:00
}
}
#if 0
if(dmabytes) printf("dmabytes %d ", dmabytes);
if(piobytes) printf("piobytes %d", piobytes);
if(dmabytes || piobytes) printf("\n");
#endif
2005-04-21 16:53:53 +02:00
w_command = CMD_IDLE;
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
return(total);
2005-04-21 16:53:53 +02:00
}
/*===========================================================================*
* atapi_sendpacket *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static int atapi_sendpacket(packet, cnt, do_dma)
2005-04-21 16:53:53 +02:00
u8_t *packet;
unsigned cnt;
int do_dma;
2005-04-21 16:53:53 +02:00
{
/* Send an Atapi Packet Command */
struct wini *wn = w_wn;
pvb_pair_t outbyte[6]; /* vector for sys_voutb() */
int s;
2005-09-11 19:09:11 +02:00
if (wn->state & IGNORING) return ERR;
2005-04-21 16:53:53 +02:00
/* Select Master/Slave drive */
2005-08-25 18:25:19 +02:00
if ((s=sys_outb(wn->base_cmd + REG_DRIVE, wn->ldhpref)) != OK)
panic("Couldn't select master/ slave drive: %d", s);
2005-04-21 16:53:53 +02:00
if (!w_waitfor(STATUS_BSY | STATUS_DRQ, 0)) {
2005-05-03 10:59:13 +02:00
printf("%s: atapi_sendpacket: drive not ready\n", w_name());
2005-04-21 16:53:53 +02:00
return(ERR);
}
/* Schedule a wakeup call, some controllers are flaky. This is done with
* a synchronous alarm. If a timeout occurs a SYN_ALARM message is sent
* from HARDWARE, so that w_intr_wait() can call w_timeout() in case the
* controller was not able to execute the command. Leftover timeouts are
* simply ignored by the main loop.
*/
sys_setalarm(wakeup_ticks, 0);
2005-04-21 16:53:53 +02:00
if (cnt > 0xFFFE) cnt = 0xFFFE; /* Max data per interrupt. */
w_command = ATAPI_PACKETCMD;
pv_set(outbyte[0], wn->base_cmd + REG_FEAT, do_dma ? FEAT_DMA : 0);
2005-08-25 18:25:19 +02:00
pv_set(outbyte[1], wn->base_cmd + REG_IRR, 0);
pv_set(outbyte[2], wn->base_cmd + REG_SAMTAG, 0);
pv_set(outbyte[3], wn->base_cmd + REG_CNT_LO, (cnt >> 0) & 0xFF);
pv_set(outbyte[4], wn->base_cmd + REG_CNT_HI, (cnt >> 8) & 0xFF);
pv_set(outbyte[5], wn->base_cmd + REG_COMMAND, w_command);
2005-09-11 19:09:11 +02:00
if (atapi_debug) printf("cmd: %x ", w_command);
2005-04-21 16:53:53 +02:00
if ((s=sys_voutb(outbyte,6)) != OK)
panic("Couldn't write registers with sys_voutb(): %d", s);
2005-04-21 16:53:53 +02:00
if (!w_waitfor(STATUS_BSY | STATUS_DRQ, STATUS_DRQ)) {
2005-08-25 15:14:02 +02:00
printf("%s: timeout (BSY|DRQ -> DRQ)\n", w_name());
2005-04-21 16:53:53 +02:00
return(ERR);
}
wn->w_status |= STATUS_ADMBSY; /* Command not at all done yet. */
2005-04-21 16:53:53 +02:00
/* Send the command packet to the device. */
if ((s=sys_outsw(wn->base_cmd + REG_DATA, SELF, packet, ATAPI_PACKETSIZE)) != OK)
panic("sys_outsw() failed: %d", s);
2005-04-21 16:53:53 +02:00
return(OK);
}
2005-09-11 19:09:11 +02:00
/*===========================================================================*
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
* w_ioctl *
2005-09-11 19:09:11 +02:00
*===========================================================================*/
static int w_ioctl(devminor_t minor, unsigned long request, endpoint_t endpt,
cp_grant_id_t grant, endpoint_t UNUSED(user_endpt))
{
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
int r, timeout, prev, count;
2010-08-12 16:09:34 +02:00
struct command cmd;
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
switch (request) {
case DIOCTIMEOUT:
r= sys_safecopyfrom(endpt, grant, 0, (vir_bytes)&timeout,
sizeof(timeout));
if(r != OK)
return r;
2005-09-11 19:09:11 +02:00
if (timeout == 0) {
/* Restore defaults. */
timeout_usecs = DEF_TIMEOUT_USECS;
max_errors = MAX_ERRORS;
wakeup_ticks = WAKEUP_TICKS;
w_silent = 0;
2005-09-11 19:09:11 +02:00
} else if (timeout < 0) {
return EINVAL;
} else {
prev = wakeup_ticks;
2005-09-11 19:09:11 +02:00
if (!w_standard_timeouts) {
/* Set (lower) timeout, lower error
* tolerance and set silent mode.
*/
wakeup_ticks = timeout;
max_errors = 3;
w_silent = 1;
timeout = timeout * 1000000 / sys_hz();
if (timeout_usecs > timeout)
timeout_usecs = timeout;
}
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
r= sys_safecopyto(endpt, grant, 0, (vir_bytes)&prev,
sizeof(prev));
if(r != OK)
return r;
}
return OK;
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
case DIOCOPENCT:
if (w_prepare(minor) == NULL) return ENXIO;
count = w_wn->open_ct;
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
r= sys_safecopyto(endpt, grant, 0, (vir_bytes)&count,
sizeof(count));
if(r != OK)
2005-08-05 15:50:50 +02:00
return r;
return OK;
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
case DIOCFLUSH:
if (w_prepare(minor) == NULL) return ENXIO;
2010-08-12 16:09:34 +02:00
if (w_wn->state & ATAPI) return EINVAL;
if (!(w_wn->state & INITIALIZED) && w_specify() != OK)
return EIO;
cmd.command = CMD_FLUSH_CACHE;
if (com_simple(&cmd) != OK || !w_waitfor(STATUS_BSY, 0))
return EIO;
return (w_wn->w_status & (STATUS_ERR|STATUS_WF)) ? EIO : OK;
2005-08-05 15:50:50 +02:00
}
Split block/character protocols and libdriver This patch separates the character and block driver communication protocols. The old character protocol remains the same, but a new block protocol is introduced. The libdriver library is replaced by two new libraries: libchardriver and libblockdriver. Their exposed API, and drivers that use them, have been updated accordingly. Together, libbdev and libblockdriver now completely abstract away the message format used by the block protocol. As the memory driver is both a character and a block device driver, it now implements its own message loop. The most important semantic change made to the block protocol is that it is no longer possible to return both partial results and an error for a single transfer. This simplifies the interaction between the caller and the driver, as the I/O vector no longer needs to be copied back. Also, drivers are now no longer supposed to decide based on the layout of the I/O vector when a transfer should be cut short. Put simply, transfers are now supposed to either succeed completely, or result in an error. After this patch, the state of the various pieces is as follows: - block protocol: stable - libbdev API: stable for synchronous communication - libblockdriver API: needs slight revision (the drvlib/partition API in particular; the threading API will also change shortly) - character protocol: needs cleanup - libchardriver API: needs cleanup accordingly - driver restarts: largely unsupported until endpoint changes are reintroduced As a side effect, this patch eliminates several bugs, hacks, and gcc -Wall and -W warnings all over the place. It probably introduces a few new ones, too. Update warning: this patch changes the protocol between MFS and disk drivers, so in order to use old/new images, the MFS from the ramdisk must be used to mount all file systems.
2011-11-22 13:27:53 +01:00
return ENOTTY;
}
2005-09-11 19:09:11 +02:00
/*===========================================================================*
* w_hw_int *
*===========================================================================*/
static void w_hw_int(unsigned int UNUSED(irqs))
{
/* Leftover interrupt(s) received; ack it/them. For native drives only. */
unsigned int drive;
2012-03-05 00:11:41 +01:00
u32_t w_status;
2006-02-24 13:55:47 +01:00
for (drive = 0; drive < MAX_DRIVES; drive++) {
if (!(wini[drive].state & IGNORING) && wini[drive].native) {
2006-02-24 13:55:47 +01:00
if (sys_inb((wini[drive].base_cmd + REG_STATUS),
&w_status) != OK)
{
panic("couldn't ack irq on drive: %d", drive);
2006-02-24 13:55:47 +01:00
}
wini[drive].w_status= w_status;
sys_inb(wini[drive].base_dma + DMA_STATUS, &w_status);
if(w_status & DMA_ST_INT) {
sys_outb(wini[drive].base_dma + DMA_STATUS, DMA_ST_INT);
wini[drive].dma_intseen = 1;
}
if (sys_irqenable(&wini[drive].irq_hook_id) != OK)
printf("couldn't re-enable drive %d\n", drive);
}
}
}
2012-07-27 16:49:55 +02:00
#define STSTR(a) if (status & STATUS_ ## a) strlcat(str, #a " ", sizeof(str));
#define ERRSTR(a) if (e & ERROR_ ## a) strlcat(str, #a " ", sizeof(str));
2012-03-25 20:25:53 +02:00
static char *strstatus(int status)
{
static char str[200];
str[0] = '\0';
STSTR(BSY);
STSTR(DRDY);
STSTR(DMADF);
STSTR(SRVCDSC);
STSTR(DRQ);
STSTR(CORR);
STSTR(CHECK);
return str;
}
2012-03-25 20:25:53 +02:00
static char *strerr(int e)
{
static char str[200];
str[0] = '\0';
ERRSTR(BB);
ERRSTR(ECC);
ERRSTR(ID);
ERRSTR(AC);
ERRSTR(TK);
ERRSTR(DM);
return str;
}
2005-09-11 19:09:11 +02:00
/*===========================================================================*
* atapi_intr_wait *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
static int atapi_intr_wait(int UNUSED(do_dma), size_t UNUSED(max))
2005-04-21 16:53:53 +02:00
{
/* Wait for an interrupt and study the results. Returns a number of bytes
* that need to be transferred, or an error code.
*/
struct wini *wn = w_wn;
pvb_pair_t inbyte[4]; /* vector for sys_vinb() */
int s; /* status for sys_vinb() */
int e;
int len;
int irr;
int r;
int phase;
w_intr_wait();
/* Request series of device I/O. */
2005-08-25 18:25:19 +02:00
inbyte[0].port = wn->base_cmd + REG_ERROR;
inbyte[1].port = wn->base_cmd + REG_CNT_LO;
inbyte[2].port = wn->base_cmd + REG_CNT_HI;
inbyte[3].port = wn->base_cmd + REG_IRR;
2005-04-21 16:53:53 +02:00
if ((s=sys_vinb(inbyte, 4)) != OK)
panic("ATAPI failed sys_vinb(): %d", s);
2005-04-21 16:53:53 +02:00
e = inbyte[0].value;
len = inbyte[1].value;
len |= inbyte[2].value << 8;
irr = inbyte[3].value;
if (wn->w_status & (STATUS_BSY | STATUS_CHECK)) {
2005-09-11 19:09:11 +02:00
if (atapi_debug) {
printf("atapi fail: S=%x=%s E=%02x=%s L=%04x I=%02x\n", wn->w_status, strstatus(wn->w_status), e, strerr(e), len, irr);
}
return ERR;
2005-04-21 16:53:53 +02:00
}
phase = (wn->w_status & STATUS_DRQ) | (irr & (IRR_COD | IRR_IO));
2005-04-21 16:53:53 +02:00
switch (phase) {
case IRR_COD | IRR_IO:
if (ATAPI_DEBUG) printf("ACD: Phase Command Complete\n");
r = OK;
break;
case 0:
if (ATAPI_DEBUG) printf("ACD: Phase Command Aborted\n");
r = ERR;
break;
case STATUS_DRQ | IRR_COD:
if (ATAPI_DEBUG) printf("ACD: Phase Command Out\n");
r = ERR;
break;
case STATUS_DRQ:
if (ATAPI_DEBUG) printf("ACD: Phase Data Out %d\n", len);
r = len;
break;
case STATUS_DRQ | IRR_IO:
if (ATAPI_DEBUG) printf("ACD: Phase Data In %d\n", len);
r = len;
break;
default:
if (ATAPI_DEBUG) printf("ACD: Phase Unknown\n");
r = ERR;
break;
}
wn->w_status |= STATUS_ADMBSY; /* Assume not done yet. */
2005-04-21 16:53:53 +02:00
return(r);
}