mirror of
https://git.openafs.org/openafs.git
synced 2025-01-19 15:30:14 +00:00
eed79e2d28
When afs_linux_read_cache is called with a non-NULL task, it is
responsible for unlocking 'page' (unless it's unlocked in a background
task), even if we encounter an error. Currently we almost always do
unlock the given page for a non-NULL task, but if we manage to hit one
of the codepaths that 'goto out', we skip over the unlock_page() call
near the end of the function, and the page never gets unlocked.
As a result, the page stays locked forever. That generally means any
future access to the same file will block forever, and when we try to
flush the relevant vcache, we will block waiting for the page lock
while holding GLOCK. (This can happen via the background daemon via
e.g. afs_ShakeLooseVCaches -> osi_TryEvictVCache -> afs_FlushVCache ->
osi_VM_FlushVCache -> vmtruncate -> ... -> truncate_inode_pages_range
-> __lock_page on Linux 2.6.32-754.2.1.el6.) This quickly brings the
whole client to a halt until the machine can be forcibly rebooted.
To solve this, just move the 'out:' label to before the page unlock.
Add a few locking-related comments around the relevant code to help
explain some relevant details.
The relevant code has changed and been refactored over the years, but
this problem has probably existed ever since this code was originally
converted to using the readpage() of the underlying cache fs, in
commit
|
||
|---|---|---|
| build-tools | ||
| doc | ||
| src | ||
| tests | ||
| .gitignore | ||
| .gitreview | ||
| .mailmap | ||
| .splintrc | ||
| acinclude.m4 | ||
| CODING | ||
| configure-libafs.ac | ||
| configure.ac | ||
| CONTRIBUTING | ||
| INSTALL | ||
| libafsdep | ||
| LICENSE | ||
| Makefile-libafs.in | ||
| Makefile.in | ||
| NEWS | ||
| NTMakefile | ||
| README | ||
| README-WINDOWS | ||
| regen.sh | ||
AFS is a distributed file system that enables users to share and access all of the files stored in a network of computers as easily as they access the files stored on their local machines. The file system is called distributed for this exact reason: files can reside on many different machines, but are available to users on every machine. OpenAFS 1.0 was originally released by IBM under the terms of the IBM Public License 1.0 (IPL10). For details on IPL10 see the LICENSE file in this directory. The current OpenAFS distribution is licensed under a combination of the IPL10 and many other licenses as granted by the relevant copyright holders. The LICENSE file in this directory contains more details, thought it is not a comprehensive statement. See INSTALL for information about building and installing OpenAFS on various platforms. See CODING for developer information and guidelines. See NEWS for recent changes to OpenAFS.