minix/servers/vm/mmap.c

543 lines
13 KiB
C
Raw Normal View History

#define _SYSTEM 1
#include <minix/callnr.h>
#include <minix/com.h>
#include <minix/config.h>
#include <minix/const.h>
#include <minix/ds.h>
#include <minix/endpoint.h>
#include <minix/minlib.h>
#include <minix/type.h>
#include <minix/ipc.h>
#include <minix/sysutil.h>
#include <minix/syslib.h>
#include <minix/safecopies.h>
#include <minix/bitmap.h>
#include <minix/debug.h>
#include <sys/mman.h>
#include <sys/param.h>
#include <errno.h>
#include <assert.h>
#include <string.h>
#include <env.h>
#include <stdio.h>
#include <fcntl.h>
#include "glo.h"
#include "proto.h"
#include "util.h"
#include "region.h"
static struct vir_region *mmap_region(struct vmproc *vmp, vir_bytes addr,
u32_t vmm_flags, size_t len, u32_t vrflags,
mem_type_t *mt, int execpriv)
{
u32_t mfflags = 0;
struct vir_region *vr = NULL;
if(vmm_flags & MAP_LOWER16M) vrflags |= VR_LOWER16MB;
if(vmm_flags & MAP_LOWER1M) vrflags |= VR_LOWER1MB;
if(vmm_flags & MAP_ALIGN64K) vrflags |= VR_PHYS64K;
if(vmm_flags & MAP_PREALLOC) mfflags |= MF_PREALLOC;
if(vmm_flags & MAP_UNINITIALIZED) {
if(!execpriv) return NULL;
vrflags |= VR_UNINITIALIZED;
}
if(len <= 0) {
return NULL;
}
if(len % VM_PAGE_SIZE)
len += VM_PAGE_SIZE - (len % VM_PAGE_SIZE);
if (addr && (vmm_flags & MAP_FIXED)) {
int r = map_unmap_range(vmp, addr, len);
if(r != OK) {
printf("mmap_region: map_unmap_range failed (%d)\n", r);
return NULL;
}
}
if (addr || (vmm_flags & MAP_FIXED)) {
/* An address is given, first try at that address. */
vr = map_page_region(vmp, addr, 0, len,
vrflags, mfflags, mt);
if(!vr && (vmm_flags & MAP_FIXED))
return NULL;
}
if (!vr) {
/* No address given or address already in use. */
vr = map_page_region(vmp, VM_PAGE_SIZE, VM_DATATOP, len,
vrflags, mfflags, mt);
}
return vr;
}
static int mmap_file(struct vmproc *vmp,
int vmfd, u32_t off_lo, u32_t off_hi, int flags,
ino_t ino, dev_t dev, u64_t filesize, vir_bytes addr, vir_bytes len,
vir_bytes *retaddr, u16_t clearend, int writable, int mayclosefd)
{
/* VFS has replied to a VMVFSREQ_FDLOOKUP request. */
struct vir_region *vr;
u64_t file_offset, page_offset;
int result = OK;
u32_t vrflags = 0;
if(writable) vrflags |= VR_WRITABLE;
if(flags & MAP_THIRDPARTY) {
file_offset = off_lo;
} else {
file_offset = make64(off_lo, off_hi);
if(off_hi && !off_lo) {
/* XXX clang compatability hack */
off_hi = file_offset = 0;
}
}
/* Do some page alignments. */
if((page_offset = (file_offset % VM_PAGE_SIZE))) {
file_offset -= page_offset;
len += page_offset;
}
len = roundup(len, VM_PAGE_SIZE);
/* All numbers should be page-aligned now. */
assert(!(len % VM_PAGE_SIZE));
assert(!(filesize % VM_PAGE_SIZE));
assert(!(file_offset % VM_PAGE_SIZE));
#if 0
/* XXX ld.so relies on longer-than-file mapping */
if((u64_t) len + file_offset > filesize) {
printf("VM: truncating mmap dev 0x%x ino %d beyond file size in %d; offset %llu, len %lu, size %llu; ",
dev, ino, vmp->vm_endpoint,
file_offset, len, filesize);
len = filesize - file_offset;
return EINVAL;
}
#endif
if(!(vr = mmap_region(vmp, addr, flags, len,
vrflags, &mem_type_mappedfile, 0))) {
result = ENOMEM;
} else {
*retaddr = vr->vaddr + page_offset;
result = OK;
mappedfile_setfile(vmp, vr, vmfd,
file_offset, dev, ino, clearend, 1, mayclosefd);
}
return result;
}
int do_vfs_mmap(message *m)
{
vir_bytes v;
struct vmproc *vmp;
int r, n;
u16_t clearend, flags = 0;
/* It might be disabled */
if(!enable_filemap) return ENXIO;
clearend = (m->m_u.m_vm_vfs.clearend_and_flags & MVM_LENMASK);
flags = (m->m_u.m_vm_vfs.clearend_and_flags & MVM_FLAGSMASK);
if((r=vm_isokendpt(m->m_u.m_vm_vfs.who, &n)) != OK)
panic("bad ep %d from vfs", m->m_u.m_vm_vfs.who);
vmp = &vmproc[n];
return mmap_file(vmp, m->m_u.m_vm_vfs.fd, m->m_u.m_vm_vfs.offset, 0,
MAP_PRIVATE | MAP_FIXED,
m->m_u.m_vm_vfs.ino, m->m_u.m_vm_vfs.dev,
(u64_t) LONG_MAX * VM_PAGE_SIZE,
m->m_u.m_vm_vfs.vaddr, m->m_u.m_vm_vfs.len, &v,
clearend, flags, 0);
}
static void mmap_file_cont(struct vmproc *vmp, message *replymsg, void *cbarg,
void *origmsg_v)
{
message *origmsg = (message *) origmsg_v;
message mmap_reply;
int result;
int writable = 0;
vir_bytes v = (vir_bytes) MAP_FAILED;
if(origmsg->VMM_PROT & PROT_WRITE)
writable = 1;
if(replymsg->VMV_RESULT != OK) {
libminixfs: allow non-pagesize-multiple FSes The memory-mapped files implementation (mmap() etc.) is implemented with the help of the filesystems using the in-VM FS cache. Filesystems tell it about all cached blocks and their metadata. Metadata is: device offset and, if any (and known), inode number and in-inode offset. VM can then map in requested memory-mapped file blocks, and request them if necessary. A limitation of this system is that filesystem block sizes that are not a multiple of the VM system (and VM hardware) page size are not possible; we can't map blocks in partially. (We can copy, but then the benefits of mapping and sharing the physical pages is gone.) So until before this commit various pieces of caching code assumed page size multiple blocksizes. This isn't strictly necessary as long as mmap() needn't be supported on that FS. This change allows the in-FS cache code (libminixfs) to allocate any-sized blocks, and will not interact with the VM cache for non-pagesize-multiple blocks. In that case it will also signal requestors, by failing 'peek' requests, that mmap() should not be supported on this FS. VM and VFS will then gracefully fail all file-mapping mmap() calls, and exec() will fall back to copying executable blocks instead of mmap()ping executables. As a result, 3 diagnostics that signal file-mapped mmap()s failing (hitherto an unusual occurence) are disabled, as ld.so does file-mapped mmap()s to map in objects it needs. On FSes not supporting it this situation is legitimate and shouldn't cause so much noise. ld.so will revert to its own minix-specific allocate+copy style of starting executables if mmap()s fail. Change-Id: Iecb1c8090f5e0be28da8f5181bb35084eb18f67b
2013-11-19 16:59:52 +01:00
#if 0 /* Noisy diagnostic for mmap() by ld.so */
printf("VM: VFS reply failed (%d)\n", replymsg->VMV_RESULT);
sys_diagctl_stacktrace(vmp->vm_endpoint);
libminixfs: allow non-pagesize-multiple FSes The memory-mapped files implementation (mmap() etc.) is implemented with the help of the filesystems using the in-VM FS cache. Filesystems tell it about all cached blocks and their metadata. Metadata is: device offset and, if any (and known), inode number and in-inode offset. VM can then map in requested memory-mapped file blocks, and request them if necessary. A limitation of this system is that filesystem block sizes that are not a multiple of the VM system (and VM hardware) page size are not possible; we can't map blocks in partially. (We can copy, but then the benefits of mapping and sharing the physical pages is gone.) So until before this commit various pieces of caching code assumed page size multiple blocksizes. This isn't strictly necessary as long as mmap() needn't be supported on that FS. This change allows the in-FS cache code (libminixfs) to allocate any-sized blocks, and will not interact with the VM cache for non-pagesize-multiple blocks. In that case it will also signal requestors, by failing 'peek' requests, that mmap() should not be supported on this FS. VM and VFS will then gracefully fail all file-mapping mmap() calls, and exec() will fall back to copying executable blocks instead of mmap()ping executables. As a result, 3 diagnostics that signal file-mapped mmap()s failing (hitherto an unusual occurence) are disabled, as ld.so does file-mapped mmap()s to map in objects it needs. On FSes not supporting it this situation is legitimate and shouldn't cause so much noise. ld.so will revert to its own minix-specific allocate+copy style of starting executables if mmap()s fail. Change-Id: Iecb1c8090f5e0be28da8f5181bb35084eb18f67b
2013-11-19 16:59:52 +01:00
#endif
result = origmsg->VMV_RESULT;
} else {
/* Finish mmap */
result = mmap_file(vmp, replymsg->VMV_FD, origmsg->VMM_OFFSET_LO,
origmsg->VMM_OFFSET_HI, origmsg->VMM_FLAGS,
replymsg->VMV_INO, replymsg->VMV_DEV,
(u64_t) replymsg->VMV_SIZE_PAGES*PAGE_SIZE,
origmsg->VMM_ADDR,
origmsg->VMM_LEN, &v, 0, writable, 1);
}
/* Unblock requesting process. */
memset(&mmap_reply, 0, sizeof(mmap_reply));
mmap_reply.m_type = result;
mmap_reply.VMM_ADDR = v;
if(send(vmp->vm_endpoint, &mmap_reply) != OK)
panic("VM: mmap_file_cont: send() failed");
}
/*===========================================================================*
* do_mmap *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
int do_mmap(message *m)
{
int r, n;
struct vmproc *vmp;
vir_bytes addr = m->VMM_ADDR;
struct vir_region *vr = NULL;
int execpriv = 0;
size_t len = (vir_bytes) m->VMM_LEN;
/* RS and VFS can do slightly more special mmap() things */
if(m->m_source == VFS_PROC_NR || m->m_source == RS_PROC_NR)
execpriv = 1;
if(m->VMM_FLAGS & MAP_THIRDPARTY) {
if(!execpriv) return EPERM;
if((r=vm_isokendpt(m->VMM_FORWHOM, &n)) != OK)
return ESRCH;
} else {
/* regular mmap, i.e. for caller */
if((r=vm_isokendpt(m->m_source, &n)) != OK) {
panic("do_mmap: message from strange source: %d",
m->m_source);
}
}
vmp = &vmproc[n];
/* "SUSv3 specifies that mmap() should fail if length is 0" */
if(len <= 0) {
return EINVAL;
}
if(m->VMM_FD == -1 || (m->VMM_FLAGS & MAP_ANON)) {
/* actual memory in some form */
mem_type_t *mt = NULL;
if(m->VMM_FD != -1) {
printf("VM: mmap: fd %d, len 0x%x\n", m->VMM_FD, len);
return EINVAL;
}
/* Contiguous phys memory has to be preallocated. */
if((m->VMM_FLAGS & (MAP_CONTIG|MAP_PREALLOC)) == MAP_CONTIG) {
return EINVAL;
}
if(m->VMM_FLAGS & MAP_CONTIG) {
mt = &mem_type_anon_contig;
} else mt = &mem_type_anon;
if(!(vr = mmap_region(vmp, addr, m->VMM_FLAGS, len,
VR_WRITABLE | VR_ANON, mt, execpriv))) {
return ENOMEM;
}
} else {
/* File mapping might be disabled */
if(!enable_filemap) return ENXIO;
/* files get private copies of pages on writes. */
if(!(m->VMM_FLAGS & MAP_PRIVATE)) {
printf("VM: mmap file must MAP_PRIVATE\n");
return ENXIO;
}
if(vfs_request(VMVFSREQ_FDLOOKUP, m->VMM_FD, vmp, 0, 0,
mmap_file_cont, NULL, m, sizeof(*m)) != OK) {
printf("VM: vfs_request for mmap failed\n");
return ENXIO;
}
/* request queued; don't reply. */
return SUSPEND;
}
/* Return mapping, as seen from process. */
No more intel/minix segments. This commit removes all traces of Minix segments (the text/data/stack memory map abstraction in the kernel) and significance of Intel segments (hardware segments like CS, DS that add offsets to all addressing before page table translation). This ultimately simplifies the memory layout and addressing and makes the same layout possible on non-Intel architectures. There are only two types of addresses in the world now: virtual and physical; even the kernel and processes have the same virtual address space. Kernel and user processes can be distinguished at a glance as processes won't use 0xF0000000 and above. No static pre-allocated memory sizes exist any more. Changes to booting: . The pre_init.c leaves the kernel and modules exactly as they were left by the bootloader in physical memory . The kernel starts running using physical addressing, loaded at a fixed location given in its linker script by the bootloader. All code and data in this phase are linked to this fixed low location. . It makes a bootstrap pagetable to map itself to a fixed high location (also in linker script) and jumps to the high address. All code and data then use this high addressing. . All code/data symbols linked at the low addresses is prefixed by an objcopy step with __k_unpaged_*, so that that code cannot reference highly-linked symbols (which aren't valid yet) or vice versa (symbols that aren't valid any more). . The two addressing modes are separated in the linker script by collecting the unpaged_*.o objects and linking them with low addresses, and linking the rest high. Some objects are linked twice, once low and once high. . The bootstrap phase passes a lot of information (e.g. free memory list, physical location of the modules, etc.) using the kinfo struct. . After this bootstrap the low-linked part is freed. . The kernel maps in VM into the bootstrap page table so that VM can begin executing. Its first job is to make page tables for all other boot processes. So VM runs before RS, and RS gets a fully dynamic, VM-managed address space. VM gets its privilege info from RS as usual but that happens after RS starts running. . Both the kernel loading VM and VM organizing boot processes happen using the libexec logic. This removes the last reason for VM to still know much about exec() and vm/exec.c is gone. Further Implementation: . All segments are based at 0 and have a 4 GB limit. . The kernel is mapped in at the top of the virtual address space so as not to constrain the user processes. . Processes do not use segments from the LDT at all; there are no segments in the LDT any more, so no LLDT is needed. . The Minix segments T/D/S are gone and so none of the user-space or in-kernel copy functions use them. The copy functions use a process endpoint of NONE to realize it's a physical address, virtual otherwise. . The umap call only makes sense to translate a virtual address to a physical address now. . Segments-related calls like newmap and alloc_segments are gone. . All segments-related translation in VM is gone (vir2map etc). . Initialization in VM is simpler as no moving around is necessary. . VM and all other boot processes can be linked wherever they wish and will be mapped in at the right location by the kernel and VM respectively. Other changes: . The multiboot code is less special: it does not use mb_print for its diagnostics any more but uses printf() as normal, saving the output into the diagnostics buffer, only printing to the screen using the direct print functions if a panic() occurs. . The multiboot code uses the flexible 'free memory map list' style to receive the list of free memory if available. . The kernel determines the memory layout of the processes to a degree: it tells VM where the kernel starts and ends and where the kernel wants the top of the process to be. VM then uses this entire range, i.e. the stack is right at the top, and mmap()ped bits of memory are placed below that downwards, and the break grows upwards. Other Consequences: . Every process gets its own page table as address spaces can't be separated any more by segments. . As all segments are 0-based, there is no distinction between virtual and linear addresses, nor between userspace and kernel addresses. . Less work is done when context switching, leading to a net performance increase. (8% faster on my machine for 'make servers'.) . The layout and configuration of the GDT makes sysenter and syscall possible.
2012-05-07 16:03:35 +02:00
m->VMM_RETADDR = vr->vaddr;
return OK;
}
/*===========================================================================*
* map_perm_check *
*===========================================================================*/
static int map_perm_check(endpoint_t caller, endpoint_t target,
phys_bytes physaddr, phys_bytes len)
{
int r;
/* TTY and memory are allowed to do anything.
* They have to be special cases as they have to be able to do
* anything; TTY even on behalf of anyone for the TIOCMAPMEM
* ioctl. MEM just for itself.
*/
if(caller == TTY_PROC_NR)
return OK;
if(caller != target)
return EPERM;
if(caller == MEM_PROC_NR)
return OK;
/* Anyone else needs explicit permission from the kernel (ultimately
* set by PCI).
*/
r = sys_privquery_mem(caller, physaddr, len);
return r;
}
/*===========================================================================*
* do_map_phys *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
int do_map_phys(message *m)
{
int r, n;
struct vmproc *vmp;
endpoint_t target;
struct vir_region *vr;
vir_bytes len;
phys_bytes startaddr;
size_t offset;
target = m->VMMP_EP;
len = m->VMMP_LEN;
if (len <= 0) return EINVAL;
if(target == SELF)
target = m->m_source;
if((r=vm_isokendpt(target, &n)) != OK)
return EINVAL;
startaddr = (vir_bytes)m->VMMP_PHADDR;
/* First check permission, then round range down/up. Caller can't
* help it if we can't map in lower than page granularity.
*/
if(map_perm_check(m->m_source, target, startaddr, len) != OK) {
printf("VM: unauthorized mapping of 0x%lx by %d\n",
startaddr, m->m_source);
return EPERM;
}
vmp = &vmproc[n];
offset = startaddr % VM_PAGE_SIZE;
len += offset;
startaddr -= offset;
if(len % VM_PAGE_SIZE)
len += VM_PAGE_SIZE - (len % VM_PAGE_SIZE);
if(!(vr = map_page_region(vmp, 0, VM_DATATOP, len,
VR_DIRECT | VR_WRITABLE, 0, &mem_type_directphys))) {
return ENOMEM;
}
phys_setphys(vr, startaddr);
m->VMMP_VADDR_REPLY = (void *) (vr->vaddr + offset);
return OK;
}
/*===========================================================================*
* do_remap *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
int do_remap(message *m)
{
2010-06-09 11:14:53 +02:00
int dn, sn;
vir_bytes da, sa;
size_t size;
u32_t flags;
struct vir_region *src_region, *vr;
struct vmproc *dvmp, *svmp;
int r;
int readonly;
if(m->m_type == VM_REMAP)
readonly = 0;
else if(m->m_type == VM_REMAP_RO)
readonly = 1;
else panic("do_remap: can't be");
da = (vir_bytes) m->VMRE_DA;
sa = (vir_bytes) m->VMRE_SA;
size = m->VMRE_SIZE;
if (size <= 0) return EINVAL;
2010-06-09 11:14:53 +02:00
if ((r = vm_isokendpt((endpoint_t) m->VMRE_D, &dn)) != OK)
return EINVAL;
2010-06-09 11:14:53 +02:00
if ((r = vm_isokendpt((endpoint_t) m->VMRE_S, &sn)) != OK)
return EINVAL;
dvmp = &vmproc[dn];
svmp = &vmproc[sn];
if (!(src_region = map_lookup(svmp, sa, NULL)))
return EINVAL;
if(src_region->vaddr != sa) {
printf("VM: do_remap: not start of region.\n");
return EFAULT;
}
if (size % VM_PAGE_SIZE)
size += VM_PAGE_SIZE - size % VM_PAGE_SIZE;
if(size != src_region->length) {
printf("VM: do_remap: not size of region.\n");
return EFAULT;
}
flags = VR_SHARED;
if(!readonly)
flags |= VR_WRITABLE;
if(da)
vr = map_page_region(dvmp, da, 0, size, flags, 0,
&mem_type_shared);
else
vr = map_page_region(dvmp, 0, VM_DATATOP, size, flags, 0,
&mem_type_shared);
if(!vr) {
printf("VM: re-map of shared area failed\n");
return ENOMEM;
}
shared_setsource(vr, svmp->vm_endpoint, src_region);
m->VMRE_RETA = (char *) vr->vaddr;
return OK;
}
/*===========================================================================*
* do_get_phys *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
int do_get_phys(message *m)
{
int r, n;
struct vmproc *vmp;
endpoint_t target;
phys_bytes ret;
vir_bytes addr;
target = m->VMPHYS_ENDPT;
addr = m->VMPHYS_ADDR;
if ((r = vm_isokendpt(target, &n)) != OK)
return EINVAL;
vmp = &vmproc[n];
r = map_get_phys(vmp, addr, &ret);
m->VMPHYS_RETA = ret;
return r;
}
/*===========================================================================*
* do_get_refcount *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
int do_get_refcount(message *m)
{
int r, n;
struct vmproc *vmp;
endpoint_t target;
u8_t cnt;
vir_bytes addr;
target = m->VMREFCNT_ENDPT;
addr = m->VMREFCNT_ADDR;
if ((r = vm_isokendpt(target, &n)) != OK)
return EINVAL;
vmp = &vmproc[n];
r = map_get_ref(vmp, addr, &cnt);
m->VMREFCNT_RETC = cnt;
return r;
}
/*===========================================================================*
* do_munmap *
*===========================================================================*/
2012-03-25 20:25:53 +02:00
int do_munmap(message *m)
{
int r, n;
struct vmproc *vmp;
vir_bytes addr, len;
endpoint_t target = SELF;
if(m->m_type == VM_UNMAP_PHYS) {
target = m->VMUP_EP;
} else if(m->m_type == VM_SHM_UNMAP) {
target = m->VMUN_ENDPT;
}
if(target == SELF)
target = m->m_source;
if((r=vm_isokendpt(target, &n)) != OK) {
panic("do_mmap: message from strange source: %d", m->m_source);
}
vmp = &vmproc[n];
2012-10-13 19:07:47 +02:00
if(m->m_type == VM_UNMAP_PHYS) {
addr = (vir_bytes) m->VMUP_VADDR;
} else if(m->m_type == VM_SHM_UNMAP) {
addr = (vir_bytes) m->VMUN_ADDR;
} else addr = (vir_bytes) m->VMUM_ADDR;
if(addr % VM_PAGE_SIZE)
return EFAULT;
2012-10-13 19:07:47 +02:00
if(m->m_type == VM_UNMAP_PHYS || m->m_type == VM_SHM_UNMAP) {
struct vir_region *vr;
if(!(vr = map_lookup(vmp, addr, NULL))) {
printf("VM: unmap: address 0x%lx not found in %d\n",
addr, target);
sys_diagctl_stacktrace(target);
return EFAULT;
}
len = vr->length;
} else len = roundup(m->VMUM_LEN, VM_PAGE_SIZE);
return map_unmap_range(vmp, addr, len);
}