NTFS: Fix a mount time deadlock.

Big thanks go to Mathias Kolehmainen for reporting the bug, providing
debug output and testing the patches I sent him to get it working.

The fix was to stop calling ntfs_attr_set() at mount time as that causes
balance_dirty_pages_ratelimited() to be called which on systems with
little memory actually tries to go and balance the dirty pages which tries
to take the s_umount semaphore but because we are still in fill_super()
across which the VFS holds s_umount for writing this results in a
deadlock.

We now do the dirty work by hand by submitting individual buffers.  This
has the annoying "feature" that mounting can take a few seconds if the
journal is large as we have clear it all.  One day someone should improve
on this by deferring the journal clearing to a helper kernel thread so it
can be done in the background but I don't have time for this at the moment
and the current solution works fine so I am leaving it like this for now.

Signed-off-by: Anton Altaparmakov <aia21@cantab.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This commit is contained in:
Anton Altaparmakov 2007-10-12 09:37:15 +01:00 коммит произвёл Linus Torvalds
Родитель f26e51f67a
Коммит bfab36e816
9 изменённых файлов: 181 добавлений и 53 удалений

Просмотреть файл

@ -407,7 +407,7 @@ raiddev /dev/md0
device /dev/hda5
raid-disk 0
device /dev/hdb1
raid-disl 1
raid-disk 1
For linear raid, just change the raid-level above to "raid-level linear", for
mirrors, change it to "raid-level 1", and for stripe sets with parity, change
@ -457,6 +457,8 @@ ChangeLog
Note, a technical ChangeLog aimed at kernel hackers is in fs/ntfs/ChangeLog.
2.1.29:
- Fix a deadlock when mounting read-write.
2.1.28:
- Fix a deadlock.
2.1.27:

Просмотреть файл

@ -17,6 +17,18 @@ ToDo/Notes:
happen is unclear however so it is worth waiting until someone hits
the problem.
2.1.29 - Fix a deadlock at mount time.
- During mount the VFS holds s_umount lock on the superblock. So when
we try to empty the journal $LogFile contents by calling
ntfs_attr_set() when the machine does not have much memory and the
journal is large ntfs_attr_set() results in the VM trying to balance
dirty pages which in turn tries to that the s_umount lock and thus we
get a deadlock. The solution is to not use ntfs_attr_set() and
instead do the zeroing by hand at the block level rather than page
cache level.
- Fix sparse warnings.
2.1.28 - Fix a deadlock.
- Fix deadlock in fs/ntfs/inode.c::ntfs_put_inode(). Thanks to Sergey

Просмотреть файл

@ -6,7 +6,7 @@ ntfs-objs := aops.o attrib.o collate.o compress.o debug.o dir.o file.o \
index.o inode.o mft.o mst.o namei.o runlist.o super.o sysctl.o \
unistr.o upcase.o
EXTRA_CFLAGS = -DNTFS_VERSION=\"2.1.28\"
EXTRA_CFLAGS = -DNTFS_VERSION=\"2.1.29\"
ifeq ($(CONFIG_NTFS_DEBUG),y)
EXTRA_CFLAGS += -DDEBUG

Просмотреть файл

@ -2,7 +2,7 @@
* aops.c - NTFS kernel address space operations and page cache handling.
* Part of the Linux-NTFS project.
*
* Copyright (c) 2001-2006 Anton Altaparmakov
* Copyright (c) 2001-2007 Anton Altaparmakov
* Copyright (c) 2002 Richard Russon
*
* This program/include file is free software; you can redistribute it and/or
@ -396,7 +396,7 @@ static int ntfs_readpage(struct file *file, struct page *page)
loff_t i_size;
struct inode *vi;
ntfs_inode *ni, *base_ni;
u8 *kaddr;
u8 *addr;
ntfs_attr_search_ctx *ctx;
MFT_RECORD *mrec;
unsigned long flags;
@ -491,15 +491,15 @@ retry_readpage:
/* Race with shrinking truncate. */
attr_len = i_size;
}
kaddr = kmap_atomic(page, KM_USER0);
addr = kmap_atomic(page, KM_USER0);
/* Copy the data to the page. */
memcpy(kaddr, (u8*)ctx->attr +
memcpy(addr, (u8*)ctx->attr +
le16_to_cpu(ctx->attr->data.resident.value_offset),
attr_len);
/* Zero the remainder of the page. */
memset(kaddr + attr_len, 0, PAGE_CACHE_SIZE - attr_len);
memset(addr + attr_len, 0, PAGE_CACHE_SIZE - attr_len);
flush_dcache_page(page);
kunmap_atomic(kaddr, KM_USER0);
kunmap_atomic(addr, KM_USER0);
put_unm_err_out:
ntfs_attr_put_search_ctx(ctx);
unm_err_out:
@ -1344,7 +1344,7 @@ static int ntfs_writepage(struct page *page, struct writeback_control *wbc)
loff_t i_size;
struct inode *vi = page->mapping->host;
ntfs_inode *base_ni = NULL, *ni = NTFS_I(vi);
char *kaddr;
char *addr;
ntfs_attr_search_ctx *ctx = NULL;
MFT_RECORD *m = NULL;
u32 attr_len;
@ -1484,14 +1484,14 @@ retry_writepage:
/* Shrinking cannot fail. */
BUG_ON(err);
}
kaddr = kmap_atomic(page, KM_USER0);
addr = kmap_atomic(page, KM_USER0);
/* Copy the data from the page to the mft record. */
memcpy((u8*)ctx->attr +
le16_to_cpu(ctx->attr->data.resident.value_offset),
kaddr, attr_len);
addr, attr_len);
/* Zero out of bounds area in the page cache page. */
memset(kaddr + attr_len, 0, PAGE_CACHE_SIZE - attr_len);
kunmap_atomic(kaddr, KM_USER0);
memset(addr + attr_len, 0, PAGE_CACHE_SIZE - attr_len);
kunmap_atomic(addr, KM_USER0);
flush_dcache_page(page);
flush_dcache_mft_record_page(ctx->ntfs_ino);
/* We are done with the page. */

Просмотреть файл

@ -1,7 +1,7 @@
/**
* attrib.c - NTFS attribute operations. Part of the Linux-NTFS project.
*
* Copyright (c) 2001-2006 Anton Altaparmakov
* Copyright (c) 2001-2007 Anton Altaparmakov
* Copyright (c) 2002 Richard Russon
*
* This program/include file is free software; you can redistribute it and/or
@ -2500,7 +2500,7 @@ int ntfs_attr_set(ntfs_inode *ni, const s64 ofs, const s64 cnt, const u8 val)
struct page *page;
u8 *kaddr;
pgoff_t idx, end;
unsigned int start_ofs, end_ofs, size;
unsigned start_ofs, end_ofs, size;
ntfs_debug("Entering for ofs 0x%llx, cnt 0x%llx, val 0x%hx.",
(long long)ofs, (long long)cnt, val);
@ -2548,6 +2548,8 @@ int ntfs_attr_set(ntfs_inode *ni, const s64 ofs, const s64 cnt, const u8 val)
kunmap_atomic(kaddr, KM_USER0);
set_page_dirty(page);
page_cache_release(page);
balance_dirty_pages_ratelimited(mapping);
cond_resched();
if (idx == end)
goto done;
idx++;
@ -2604,6 +2606,8 @@ int ntfs_attr_set(ntfs_inode *ni, const s64 ofs, const s64 cnt, const u8 val)
kunmap_atomic(kaddr, KM_USER0);
set_page_dirty(page);
page_cache_release(page);
balance_dirty_pages_ratelimited(mapping);
cond_resched();
}
done:
ntfs_debug("Done.");

Просмотреть файл

@ -1,7 +1,7 @@
/*
* file.c - NTFS kernel file operations. Part of the Linux-NTFS project.
*
* Copyright (c) 2001-2006 Anton Altaparmakov
* Copyright (c) 2001-2007 Anton Altaparmakov
*
* This program/include file is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as published
@ -26,7 +26,6 @@
#include <linux/swap.h>
#include <linux/uio.h>
#include <linux/writeback.h>
#include <linux/sched.h>
#include <asm/page.h>
#include <asm/uaccess.h>
@ -362,7 +361,7 @@ static inline void ntfs_fault_in_pages_readable(const char __user *uaddr,
volatile char c;
/* Set @end to the first byte outside the last page we care about. */
end = (const char __user*)PAGE_ALIGN((ptrdiff_t __user)uaddr + bytes);
end = (const char __user*)PAGE_ALIGN((unsigned long)uaddr + bytes);
while (!__get_user(c, uaddr) && (uaddr += PAGE_SIZE, uaddr < end))
;
@ -532,7 +531,8 @@ static int ntfs_prepare_pages_for_non_resident_write(struct page **pages,
blocksize_bits = vol->sb->s_blocksize_bits;
u = 0;
do {
struct page *page = pages[u];
page = pages[u];
BUG_ON(!page);
/*
* create_empty_buffers() will create uptodate/dirty buffers if
* the page is uptodate/dirty.
@ -1291,7 +1291,7 @@ static inline size_t ntfs_copy_from_user(struct page **pages,
size_t bytes)
{
struct page **last_page = pages + nr_pages;
char *kaddr;
char *addr;
size_t total = 0;
unsigned len;
int left;
@ -1300,13 +1300,13 @@ static inline size_t ntfs_copy_from_user(struct page **pages,
len = PAGE_CACHE_SIZE - ofs;
if (len > bytes)
len = bytes;
kaddr = kmap_atomic(*pages, KM_USER0);
left = __copy_from_user_inatomic(kaddr + ofs, buf, len);
kunmap_atomic(kaddr, KM_USER0);
addr = kmap_atomic(*pages, KM_USER0);
left = __copy_from_user_inatomic(addr + ofs, buf, len);
kunmap_atomic(addr, KM_USER0);
if (unlikely(left)) {
/* Do it the slow way. */
kaddr = kmap(*pages);
left = __copy_from_user(kaddr + ofs, buf, len);
addr = kmap(*pages);
left = __copy_from_user(addr + ofs, buf, len);
kunmap(*pages);
if (unlikely(left))
goto err_out;
@ -1408,26 +1408,26 @@ static inline size_t ntfs_copy_from_user_iovec(struct page **pages,
size_t *iov_ofs, size_t bytes)
{
struct page **last_page = pages + nr_pages;
char *kaddr;
char *addr;
size_t copied, len, total = 0;
do {
len = PAGE_CACHE_SIZE - ofs;
if (len > bytes)
len = bytes;
kaddr = kmap_atomic(*pages, KM_USER0);
copied = __ntfs_copy_from_user_iovec_inatomic(kaddr + ofs,
addr = kmap_atomic(*pages, KM_USER0);
copied = __ntfs_copy_from_user_iovec_inatomic(addr + ofs,
*iov, *iov_ofs, len);
kunmap_atomic(kaddr, KM_USER0);
kunmap_atomic(addr, KM_USER0);
if (unlikely(copied != len)) {
/* Do it the slow way. */
kaddr = kmap(*pages);
copied = __ntfs_copy_from_user_iovec_inatomic(kaddr + ofs,
addr = kmap(*pages);
copied = __ntfs_copy_from_user_iovec_inatomic(addr + ofs,
*iov, *iov_ofs, len);
/*
* Zero the rest of the target like __copy_from_user().
*/
memset(kaddr + ofs + copied, 0, len - copied);
memset(addr + ofs + copied, 0, len - copied);
kunmap(*pages);
if (unlikely(copied != len))
goto err_out;
@ -1735,8 +1735,6 @@ static int ntfs_commit_pages_after_write(struct page **pages,
read_unlock_irqrestore(&ni->size_lock, flags);
BUG_ON(initialized_size != i_size);
if (end > initialized_size) {
unsigned long flags;
write_lock_irqsave(&ni->size_lock, flags);
ni->initialized_size = end;
i_size_write(vi, end);

Просмотреть файл

@ -34,7 +34,6 @@
#include "dir.h"
#include "debug.h"
#include "inode.h"
#include "attrib.h"
#include "lcnalloc.h"
#include "malloc.h"
#include "mft.h"
@ -2500,8 +2499,6 @@ retry_truncate:
/* Resize the attribute record to best fit the new attribute size. */
if (new_size < vol->mft_record_size &&
!ntfs_resident_attr_value_resize(m, a, new_size)) {
unsigned long flags;
/* The resize succeeded! */
flush_dcache_mft_record_page(ctx->ntfs_ino);
mark_mft_record_dirty(ctx->ntfs_ino);

Просмотреть файл

@ -1,7 +1,7 @@
/*
* logfile.c - NTFS kernel journal handling. Part of the Linux-NTFS project.
*
* Copyright (c) 2002-2005 Anton Altaparmakov
* Copyright (c) 2002-2007 Anton Altaparmakov
*
* This program/include file is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as published
@ -724,24 +724,139 @@ bool ntfs_is_logfile_clean(struct inode *log_vi, const RESTART_PAGE_HEADER *rp)
*/
bool ntfs_empty_logfile(struct inode *log_vi)
{
ntfs_volume *vol = NTFS_SB(log_vi->i_sb);
VCN vcn, end_vcn;
ntfs_inode *log_ni = NTFS_I(log_vi);
ntfs_volume *vol = log_ni->vol;
struct super_block *sb = vol->sb;
runlist_element *rl;
unsigned long flags;
unsigned block_size, block_size_bits;
int err;
bool should_wait = true;
ntfs_debug("Entering.");
if (!NVolLogFileEmpty(vol)) {
int err;
err = ntfs_attr_set(NTFS_I(log_vi), 0, i_size_read(log_vi),
0xff);
if (unlikely(err)) {
ntfs_error(vol->sb, "Failed to fill $LogFile with "
"0xff bytes (error code %i).", err);
return false;
}
/* Set the flag so we do not have to do it again on remount. */
NVolSetLogFileEmpty(vol);
if (NVolLogFileEmpty(vol)) {
ntfs_debug("Done.");
return true;
}
/*
* We cannot use ntfs_attr_set() because we may be still in the middle
* of a mount operation. Thus we do the emptying by hand by first
* zapping the page cache pages for the $LogFile/$DATA attribute and
* then emptying each of the buffers in each of the clusters specified
* by the runlist by hand.
*/
block_size = sb->s_blocksize;
block_size_bits = sb->s_blocksize_bits;
vcn = 0;
read_lock_irqsave(&log_ni->size_lock, flags);
end_vcn = (log_ni->initialized_size + vol->cluster_size_mask) >>
vol->cluster_size_bits;
read_unlock_irqrestore(&log_ni->size_lock, flags);
truncate_inode_pages(log_vi->i_mapping, 0);
down_write(&log_ni->runlist.lock);
rl = log_ni->runlist.rl;
if (unlikely(!rl || vcn < rl->vcn || !rl->length)) {
map_vcn:
err = ntfs_map_runlist_nolock(log_ni, vcn, NULL);
if (err) {
ntfs_error(sb, "Failed to map runlist fragment (error "
"%d).", -err);
goto err;
}
rl = log_ni->runlist.rl;
BUG_ON(!rl || vcn < rl->vcn || !rl->length);
}
/* Seek to the runlist element containing @vcn. */
while (rl->length && vcn >= rl[1].vcn)
rl++;
do {
LCN lcn;
sector_t block, end_block;
s64 len;
/*
* If this run is not mapped map it now and start again as the
* runlist will have been updated.
*/
lcn = rl->lcn;
if (unlikely(lcn == LCN_RL_NOT_MAPPED)) {
vcn = rl->vcn;
goto map_vcn;
}
/* If this run is not valid abort with an error. */
if (unlikely(!rl->length || lcn < LCN_HOLE))
goto rl_err;
/* Skip holes. */
if (lcn == LCN_HOLE)
continue;
block = lcn << vol->cluster_size_bits >> block_size_bits;
len = rl->length;
if (rl[1].vcn > end_vcn)
len = end_vcn - rl->vcn;
end_block = (lcn + len) << vol->cluster_size_bits >>
block_size_bits;
/* Iterate over the blocks in the run and empty them. */
do {
struct buffer_head *bh;
/* Obtain the buffer, possibly not uptodate. */
bh = sb_getblk(sb, block);
BUG_ON(!bh);
/* Setup buffer i/o submission. */
lock_buffer(bh);
bh->b_end_io = end_buffer_write_sync;
get_bh(bh);
/* Set the entire contents of the buffer to 0xff. */
memset(bh->b_data, -1, block_size);
if (!buffer_uptodate(bh))
set_buffer_uptodate(bh);
if (buffer_dirty(bh))
clear_buffer_dirty(bh);
/*
* Submit the buffer and wait for i/o to complete but
* only for the first buffer so we do not miss really
* serious i/o errors. Once the first buffer has
* completed ignore errors afterwards as we can assume
* that if one buffer worked all of them will work.
*/
submit_bh(WRITE, bh);
if (should_wait) {
should_wait = false;
wait_on_buffer(bh);
if (unlikely(!buffer_uptodate(bh)))
goto io_err;
}
brelse(bh);
} while (++block < end_block);
} while ((++rl)->vcn < end_vcn);
up_write(&log_ni->runlist.lock);
/*
* Zap the pages again just in case any got instantiated whilst we were
* emptying the blocks by hand. FIXME: We may not have completed
* writing to all the buffer heads yet so this may happen too early.
* We really should use a kernel thread to do the emptying
* asynchronously and then we can also set the volume dirty and output
* an error message if emptying should fail.
*/
truncate_inode_pages(log_vi->i_mapping, 0);
/* Set the flag so we do not have to do it again on remount. */
NVolSetLogFileEmpty(vol);
ntfs_debug("Done.");
return true;
io_err:
ntfs_error(sb, "Failed to write buffer. Unmount and run chkdsk.");
goto dirty_err;
rl_err:
ntfs_error(sb, "Runlist is corrupt. Unmount and run chkdsk.");
dirty_err:
NVolSetErrors(vol);
err = -EIO;
err:
up_write(&log_ni->runlist.lock);
ntfs_error(sb, "Failed to fill $LogFile with 0xff bytes (error %d).",
-err);
return false;
}
#endif /* NTFS_RW */

Просмотреть файл

@ -1,7 +1,7 @@
/**
* runlist.c - NTFS runlist handling code. Part of the Linux-NTFS project.
*
* Copyright (c) 2001-2005 Anton Altaparmakov
* Copyright (c) 2001-2007 Anton Altaparmakov
* Copyright (c) 2002-2005 Richard Russon
*
* This program/include file is free software; you can redistribute it and/or
@ -1714,7 +1714,7 @@ extend_hole:
sizeof(*rl));
/* Adjust the beginning of the tail if necessary. */
if (end > rl->vcn) {
s64 delta = end - rl->vcn;
delta = end - rl->vcn;
rl->vcn = end;
rl->length -= delta;
/* Only adjust the lcn if it is real. */