Sasha, Also wanted to make sure stable gets this patch from David Howells that fixes a regression that was mentioned in some of the same email threads. It fixes an important stable regression in cifs_readv when cache=none. Bharath has also reviewed and tested it with 6.6 stable. See attached.

On Fri, Nov 14, 2025 at 9:00 AM Sasha Levin sashal@kernel.org wrote:

On Fri, Nov 14, 2025 at 04:42:57PM +0530, Shyam Prasad N wrote:

...

Hi Greg/Sasha,

Over the last few months, a few users have reported a data corruption with Linux SMB kernel client filesystem. This is one such report: https://lore.kernel.org/linux-cifs/36fb31bf2c854cdc930a3415f5551dcd@izw-berl...

The issue is now well understood. Attached is a fix for this issue. I've made sure that the fix is stopping the data corruption and also not regressing other write patterns.

The regression seems to have been introduced during a refactoring of this code path during the v6.3 and continued to exist till v6.9, before the code was refactored again with netfs helper library integration in v6.10.

I request you to include this change in all stable trees for v6.3..v6.9. I've done my testing on stable-6.6. Please let me know if you want this tested on any other kernels.

I'll queue it up for 6.6, thanks!

-- Thanks, Sasha

-- Thanks,

Steve