cache=strict in 3.6.7 works fine, but 3.7-rc* doesn't. With cache=none 3.7-rc7 works fine. maxim ~ $ uname -a Linux maxim 3.6.7 #1 SMP PREEMPT Wed Nov 21 15:11:28 FET 2012 x86_64 AMD Athlon(tm) II X2 220 Processor AuthenticAMD GNU/Linux maxim ~ $ wget --no-proxy ftp://ftp.mozilla.org/pub/mozilla.org/thunderbird/releases/17.0/win32/ru/Thunderbird%20Setup%2017.0.exe -O /mnt/install/delme.7z maxim ~ $ gpg --print-md md5 /mnt/install/delme.7z /mnt/install/delme.7z: C2 EE 39 93 C6 38 DE AA 6B B8 7D 31 9C 31 E2 09 maxim ~ $ 7z t /mnt/install/delme.7z >/dev/null ; echo $? 0 maxim ~ $ rm /mnt/install/delme.7z maxim ~ $ uname -a Linux maxim 3.7.0-rc7 #1 SMP PREEMPT Mon Nov 26 10:43:42 FET 2012 x86_64 AMD Athlon(tm) II X2 220 Processor AuthenticAMD GNU/Linux maxim ~ $ wget --no-proxy ftp://ftp.mozilla.org/pub/mozilla.org/thunderbird/releases/17.0/win32/ru/Thunderbird%20Setup%2017.0.exe -O /mnt/install/delme.7z maxim ~ $ wget --no-proxy ftp://ftp.mozilla.org/pub/mozilla.org/thunderbird/releases/17.0/win32/ru/Thunderbird%20Setup%2017.0.exe -O /tmp/delme2.7z maxim ~ $ cp /tmp/delme2.7z /mnt/install/ maxim ~ $ gpg --print-md md5 /mnt/install/delme.7z /mnt/install/delme.7z: F9 FB 8C 09 23 98 DF EB 96 95 47 D4 93 97 26 A1 maxim ~ $ gpg --print-md md5 /mnt/install/delme2.7z /mnt/install/delme2.7z: C2 EE 39 93 C6 38 DE AA 6B B8 7D 31 9C 31 E2 09 any add info from me? windows 2003 server on another side (checked with win7 too).
Thanks for the bug report. I've been able to reproduce it... While writing out the file, the kernel is occasionally issuing a short write to the server and then is skipping over the gap and writing out the rest of the file. The gaps are not page aligned, so I'm at a bit of a loss for the cause so far...
Created attachment 87291 [details] cifs: fix writeback race with file that is growing Maxim, this patch seems to fix it for me. Does it also fix it for you? I'll plan to send this on to the list tomorrow unless you say it doesn't. With luck, we'll get it in before 3.7 ships.
Thank you! Works fine now for me
Thanks. Patch sent to maintainer and linux-cifs ml: http://article.gmane.org/gmane.linux.kernel.cifs/7414
Plan to request upstream merge after one more day in cifs-2.6.git (so it goes through linux-next first)
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=086486e46e4206cfa1140fb9682ad67c8a4502fb
Yep, merged last night. Closing bug.
A patch referencing this bug report has been merged in Linux v3.7-rc8: commit 3a98b8614312026d489e56c1d0e294a68e2aad77 Author: Jeff Layton <jlayton@redhat.com> Date: Mon Nov 26 09:48:41 2012 -0500 cifs: fix writeback race with file that is growing