The patch titled
ecryptfs: fix fsx data corruption problems
has been added to the -mm tree. Its filename is
ecryptfs-fix-fsx-data-corruption-problems.patch
*** Remember to use Documentation/SubmitChecklist when testing your code ***
See http://www.zip.com.au/~akpm/linux/patches/stuff/added-to-mm.txt to find
out what to do about this
------------------------------------------------------
Subject: ecryptfs: fix fsx data corruption problems
From: Eric Sandeen <sandeen@xxxxxxxxxx>
ecryptfs in 2.6.24-rc3 wasn't surviving fsx for me at all, dying after 4
ops. Generally, encountering problems with stale data and improperly
zeroed pages. An extending truncate + write for example would expose stale
data.
With the changes below I got to a million ops and beyond with all mmap ops
disabled - mmap still needs work. (A version of this patch on a RHEL5
kernel ran for over 110 million fsx ops)
I added a few comments as well, to the best of my understanding
as I read through the code.
Signed-off-by: Eric Sandeen <sandeen@xxxxxxxxxx>
Cc: Michael Halcrow <mhalcrow@xxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
fs/ecryptfs/mmap.c | 31 ++++++++++++++++++++-----------
fs/ecryptfs/read_write.c | 27 +++++++++++++++++++++------
2 files changed, 41 insertions(+), 17 deletions(-)
diff -puN fs/ecryptfs/mmap.c~ecryptfs-fix-fsx-data-corruption-problems fs/ecryptfs/mmap.c
--- a/fs/ecryptfs/mmap.c~ecryptfs-fix-fsx-data-corruption-problems
+++ a/fs/ecryptfs/mmap.c
@@ -263,14 +263,13 @@ out:
return 0;
}
+/* This function must zero any hole we create */
static int ecryptfs_prepare_write(struct file *file, struct page *page,
unsigned from, unsigned to)
{
int rc = 0;
+ loff_t prev_page_end_size;
- if (from == 0 && to == PAGE_CACHE_SIZE)
- goto out; /* If we are writing a full page, it will be
- up to date. */
if (!PageUptodate(page)) {
rc = ecryptfs_read_lower_page_segment(page, page->index, 0,
PAGE_CACHE_SIZE,
@@ -283,22 +282,32 @@ static int ecryptfs_prepare_write(struct
} else
SetPageUptodate(page);
}
- if (page->index != 0) {
- loff_t end_of_prev_pg_pos =
- (((loff_t)page->index << PAGE_CACHE_SHIFT) - 1);
- if (end_of_prev_pg_pos > i_size_read(page->mapping->host)) {
+ prev_page_end_size = ((loff_t)page->index << PAGE_CACHE_SHIFT);
+
+ /*
+ * If creating a page or more of holes, zero them out via truncate.
+ * Note, this will increase i_size.
+ */
+ if (page->index != 0) {
+ if (prev_page_end_size > i_size_read(page->mapping->host)) {
rc = ecryptfs_truncate(file->f_path.dentry,
- end_of_prev_pg_pos);
+ prev_page_end_size);
if (rc) {
printk(KERN_ERR "Error on attempt to "
"truncate to (higher) offset [%lld];"
- " rc = [%d]\n", end_of_prev_pg_pos, rc);
+ " rc = [%d]\n", prev_page_end_size, rc);
goto out;
}
}
- if (end_of_prev_pg_pos + 1 > i_size_read(page->mapping->host))
- zero_user_page(page, 0, PAGE_CACHE_SIZE, KM_USER0);
+ }
+ /*
+ * Writing to a new page, and creating a small hole from start of page?
+ * Zero it out.
+ */
+ if ((i_size_read(page->mapping->host) == prev_page_end_size) &&
+ (from != 0)) {
+ zero_user_page(page, 0, PAGE_CACHE_SIZE, KM_USER0);
}
out:
return rc;
diff -puN fs/ecryptfs/read_write.c~ecryptfs-fix-fsx-data-corruption-problems fs/ecryptfs/read_write.c
--- a/fs/ecryptfs/read_write.c~ecryptfs-fix-fsx-data-corruption-problems
+++ a/fs/ecryptfs/read_write.c
@@ -124,6 +124,10 @@ int ecryptfs_write(struct file *ecryptfs
loff_t pos;
int rc = 0;
+ /*
+ * if we are writing beyond current size, then start pos
+ * at the current size - we'll fill in zeros from there.
+ */
if (offset > ecryptfs_file_size)
pos = ecryptfs_file_size;
else
@@ -137,6 +141,7 @@ int ecryptfs_write(struct file *ecryptfs
if (num_bytes > total_remaining_bytes)
num_bytes = total_remaining_bytes;
if (pos < offset) {
+ /* remaining zeros to write, up to destination offset */
size_t total_remaining_zeros = (offset - pos);
if (num_bytes > total_remaining_zeros)
@@ -167,17 +172,27 @@ int ecryptfs_write(struct file *ecryptfs
}
}
ecryptfs_page_virt = kmap_atomic(ecryptfs_page, KM_USER0);
+
+ /*
+ * pos: where we're now writing, offset: where the request was
+ * If current pos is before request, we are filling zeros
+ * If we are at or beyond request, we are writing the *data*
+ * If we're in a fresh page beyond eof, zero it in either case
+ */
+ if (pos < offset || !start_offset_in_page) {
+ /* We are extending past the previous end of the file.
+ * Fill in zero values to the end of the page */
+ memset(((char *)ecryptfs_page_virt
+ + start_offset_in_page), 0,
+ PAGE_CACHE_SIZE - start_offset_in_page);
+ }
+
+ /* pos >= offset, we are now writing the data request */
if (pos >= offset) {
memcpy(((char *)ecryptfs_page_virt
+ start_offset_in_page),
(data + data_offset), num_bytes);
data_offset += num_bytes;
- } else {
- /* We are extending past the previous end of the file.
- * Fill in zero values up to the start of where we
- * will be writing data. */
- memset(((char *)ecryptfs_page_virt
- + start_offset_in_page), 0, num_bytes);
}
kunmap_atomic(ecryptfs_page_virt, KM_USER0);
flush_dcache_page(ecryptfs_page);
_
Patches currently in -mm which might be from sandeen@xxxxxxxxxx are
ecryptfs-initialize-new-auth_tokens-before-teardown.patch
ecryptfs-set-s_blocksize-from-lower-fs-in-sb.patch
ecryptfs-fix-fsx-data-corruption-problems.patch
ext2-change-the-default-behaviour-on-error.patch
ecryptfs-make-show_options-reflect-actual-mount-options.patch
ecryptfs-make-show_options-reflect-actual-mount-options-fix.patch
ext4-fix-mb_debug-format-warnings.patch
ext4-fix-freespace-accounting-with-mballoc-on-32bit-machines.patch
ext4-fix-oops-with-jbd-stats-through-procfs-and-external.patch
ext4-fix-up-ext4fs_debug-builds.patch
ext3-change-the-default-behaviour-on-error.patch
-
To unsubscribe from this list: send the line "unsubscribe mm-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
