linux.git/fs/coda, branch v5.17

coda: bump module version to 7.2

2021-11-09T18:02:51+00:00

Helps with tracking which patches have been propagated upstream and if
users are running the latest known version.

Link: https://lkml.kernel.org/r/20210908140308.18491-10-jaharkes@cs.cmu.edu
Signed-off-by: Jan Harkes 
Cc: Alex Shi 
Cc: Jing Yangyang 
Cc: Xin Tan 
Cc: Xiyu Yang 
Cc: Zeal Robot 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

coda: use vmemdup_user to replace the open code

2021-11-09T18:02:51+00:00

vmemdup_user is better than duplicating its implementation, So just
replace the open code.

  fs/coda/psdev.c:125:10-18:WARNING:opportunity for vmemdup_user

The issue is detected with the help of Coccinelle.

Link: https://lkml.kernel.org/r/20210908140308.18491-9-jaharkes@cs.cmu.edu
Reported-by: Zeal Robot 
Signed-off-by: Jing Yangyang 
Signed-off-by: Jan Harkes 
Cc: Alex Shi 
Cc: Xin Tan 
Cc: Xiyu Yang 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

coda: convert from atomic_t to refcount_t on coda_vm_ops->refcnt

2021-11-09T18:02:51+00:00

refcount_t type and corresponding API can protect refcounters from
accidental underflow and overflow and further use-after-free situations.

Link: https://lkml.kernel.org/r/20210908140308.18491-8-jaharkes@cs.cmu.edu
Signed-off-by: Xiyu Yang 
Signed-off-by: Xin Tan 
Signed-off-by: Jan Harkes 
Cc: Alex Shi 
Cc: Jing Yangyang 
Cc: Zeal Robot 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

coda: avoid doing bad things on inode type changes during revalidation

2021-11-09T18:02:51+00:00

When Coda discovers an inconsistent object, it turns it into a symlink.
However we can't just follow this change in the kernel on an existing file
or directory inode that may still have references.

This patch removes the inconsistent inode from the inode hash and
allocates a new inode for the symlink object.

Link: https://lkml.kernel.org/r/20210908140308.18491-7-jaharkes@cs.cmu.edu
Signed-off-by: Jan Harkes 
Cc: Alex Shi 
Cc: Jing Yangyang 
Cc: Xin Tan 
Cc: Xiyu Yang 
Cc: Zeal Robot 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

coda: avoid hidden code duplication in rename

2021-11-09T18:02:51+00:00

We were actually fixing up the directory mtime in both branches after the
negative dentry test, it was just that one branch was only flagging the
directory inodes to refresh their attributes while the other branch used
the optional optimization to set mtime to the current time and not go back
to the Coda client.

Link: https://lkml.kernel.org/r/20210908140308.18491-6-jaharkes@cs.cmu.edu
Signed-off-by: Jan Harkes 
Cc: Alex Shi 
Cc: Jing Yangyang 
Cc: Xin Tan 
Cc: Xiyu Yang 
Cc: Zeal Robot 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

coda: avoid flagging NULL inodes

2021-11-09T18:02:51+00:00

Somehow we hit a negative dentry in coda_rename even after checking with
d_really_is_positive.  Maybe something raced and turned the new_dentry
negative while we were fixing up directory link counts.

Link: https://lkml.kernel.org/r/20210908140308.18491-5-jaharkes@cs.cmu.edu
Signed-off-by: Jan Harkes 
Cc: Alex Shi 
Cc: Jing Yangyang 
Cc: Xin Tan 
Cc: Xiyu Yang 
Cc: Zeal Robot 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

coda: remove err which no one care

2021-11-09T18:02:51+00:00

No one care 'err' in func coda_release, so better remove it.

Link: https://lkml.kernel.org/r/20210908140308.18491-4-jaharkes@cs.cmu.edu
Signed-off-by: Alex Shi 
Signed-off-by: Jan Harkes 
Cc: Jing Yangyang 
Cc: Xin Tan 
Cc: Xiyu Yang 
Cc: Zeal Robot 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

coda: check for async upcall request using local state

2021-11-09T18:02:51+00:00

Originally flagged by Smatch because the code implicitly assumed outSize
is not NULL for non-async upcalls because of a flag that was (not) set in
req->uc_flags.

However req->uc_flags field is in shared state and although the current
code will not allow it to be changed before the async request check the
code is more robust when it tests against the local outSize variable.

Link: https://lkml.kernel.org/r/20210908140308.18491-3-jaharkes@cs.cmu.edu
Signed-off-by: Jan Harkes 
Cc: Alex Shi 
Cc: Jing Yangyang 
Cc: Xin Tan 
Cc: Xiyu Yang 
Cc: Zeal Robot 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

coda: avoid NULL pointer dereference from a bad inode

2021-11-09T18:02:51+00:00

Patch series "Coda updates for -next".

The following patch series contains some fixes for the Coda kernel module
I've had sitting around and were tested extensively in a development
version of the Coda kernel module that lives outside of the main kernel.

This patch (of 9):

Avoid accessing coda_inode_info from a dentry with a bad inode.

Link: https://lkml.kernel.org/r/20210908140308.18491-1-jaharkes@cs.cmu.edu
Link: https://lkml.kernel.org/r/20210908140308.18491-2-jaharkes@cs.cmu.edu
Signed-off-by: Jan Harkes 
Cc: Alex Shi 
Cc: Jing Yangyang 
Cc: Xin Tan 
Cc: Xiyu Yang 
Cc: Zeal Robot 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

coda: fix reference counting in coda_file_mmap error path

2021-04-23T21:42:39+00:00

mmap_region() now calls fput() on the vma->vm_file.

So we need to drop the extra reference on the coda file instead of the
host file.

Link: https://lkml.kernel.org/r/20210421132012.82354-1-christian.koenig@amd.com
Fixes: 1527f926fd04 ("mm: mmap: fix fput in error path v2")
Signed-off-by: Christian König 
Reviewed-by: Daniel Vetter 
Acked-by: Jan Harkes 
Cc: Miklos Szeredi 
Cc: Jason Gunthorpe 
Cc: 	[5.11+]
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds