Bug 6823 - Lock test behaviour is different and wrong
Summary: Lock test behaviour is different and wrong
Status: CLOSED FIXED
Alias: None
Product: CifsVFS
Classification: Unclassified
Component: kernel fs (show other bugs)
Version: 2.6
Hardware: Other Linux
: P3 normal
Target Milestone: ---
Assignee: Steve French
QA Contact:
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2009-10-18 00:58 UTC by Pavel Shilovsky
Modified: 2011-02-25 03:52 UTC (History)
0 users

See Also:


Attachments
The test, that shows the problem (6.87 KB, text/x-csrc)
2009-10-18 00:59 UTC, Pavel Shilovsky
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description Pavel Shilovsky 2009-10-18 00:58:06 UTC
This test shows problems when we mount share with forcemandatory flag.
It doesn't work: after that no one process can read from the file.

Structure of the test:
| Start first main process
| Create Unix socket
|\
| \
| | Start a first child process
| | Open a file
| | Lock it from 0 to 1
| | Open this file again
| | Lock it from 1 to 3
| | Close secondary open file descriptor
| | Send the firstly open descriptor to the main process
| | Check the area from 0 to 1 - it must be no locks there
| | Try to read this area from 0 to 1
| | Sleep 
|\
| \
| | Start a second child process process
| | Lock it from 10 to 12
| | Try to read it
| | Sleep
|  ...
|
| Receive th file descriptor from the second child process
| Check locking by first child process area 
|  (if we see lock, that prevent us - the pid is invalid - Samba Protocol doesn't allow ut to see the valid pid)
| Try to read the area from 0 to 1 
| Try to read the area from 1 to 3
| Try to read the area from 10 to 11
....


[piastry@tartarus m]$ ./a.out                                                                   
start 25123
server start 25124
server open for 5
server second open for 6
server close 6
server send 5
server file not locked
second server start 25125
fd received=5 (len=1)
second server open for 5
second server lock for 5
file 5 not locked
first read failed on 5 (Permission denied)
second lseek done on 5 with 1
second read failed on 5 (Permission denied)
third lseek done on 5 with 10
third read failed on 5 (Permission denied)
done
[piastry@tartarus m]$ second server lseek done on 5 with 10
second server read failed on 5 (Permission denied)
wait second server for end
server end

[piastry@tartarus m]$ second server end


Then I changed start position of last read from 10 to 14, but result didn't change:

[piastry@tartarus m]$ ./a.out
start 25157
server start 25158
server open for 5
server second open for 6
second server start 25159
server close 6
server send 5
fd received=5 (len=1)
server file not locked
second server open for 5
second server lock for 5
file 5 not locked
first read failed on 5 (Permission denied)
second lseek done on 5 with 1
second read failed on 5 (Permission denied)
third lseek done on 5 with 14
third read failed on 5 (Permission denied)
done
second server lseek done on 5 with 10
second server read failed on 5 (Permission denied)
wait second server for end
[piastry@tartarus m]$ server end

[piastry@tartarus m]$ second server end



After that I did 'vim "file"' and it failed with 'reading error' at first time, but at second time it worked right!

So, after that, the behavior of the test changed!!!!!!

[piastry@tartarus m]$ gcc raw7.c
raw7.c: In function ‘main’:
raw7.c:217: warning: format ‘%d’ expects type ‘int’, but argument 3 has type ‘long unsigned int’
[piastry@tartarus m]$ ./a.out
start 25811
server start 25812
server open for 5
server second open for 6
server close 6
server send 5
server file not locked
fd received=5 (len=1)
second server start 25813
second server open for 5
second server lock for 5
file 5 not locked
first read done on 5 with 1
second lseek done on 5 with 1
second read done on 5 with 2
third lseek done on 5 with 10
third read done on 5 with 2
data:ab
done
[piastry@tartarus m]$ second server lseek done on 5 with 10
second server read done on 5 with 2
wait second server for end
server end

[piastry@tartarus m]$ second server end

And with 14 instead of 10 too!

[piastry@tartarus m]$ ./a.out         
start 25732                           
server start 25733
server open for 5
server second open for 6
second server start 25734
server close 6
server send 5
server file not locked
fd received=5 (len=1)
second server open for 5
second server lock for 5
second server lseek done on 5 with 10
second server read done on 5 with 2
wait second server for end
file 5 not locked
first read done on 5 with 1
second lseek done on 5 with 1
second read done on 5 with 2
third lseek done on 5 with 14
third read done on 5 with 2
data:ef
done
[piastry@tartarus m]$ server end
second server end

[piastry@tartarus m]$ cat PDOXUSRS.NET
1234567890abcdefghghoerhjgoergoiearogeroghoerhgoerhgore
bverhbiue
ougbkelbkesrb

I use 2.6.30 kernel and CIFS Version 1.58.
All mount options: noperm,forcemand.
Comment 1 Pavel Shilovsky 2009-10-18 00:59:55 UTC
Created attachment 4863 [details]
The test, that shows the problem
Comment 2 Pavel Shilovsky 2011-02-25 03:52:23 UTC
Fixed with strictcache mount option in cifs-2.6 git tree.