Red Hat Bugzilla – Bug 59861
DOS client app reporting read/write errors against samba share
Last modified: 2007-04-18 12:40:25 EDT
From Bugzilla Helper:
User-Agent: Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; Q312461)
Description of problem:
DOS test application running on Windows 2K fails when writing to share mounted
to Linux server using Samba. The application writes and reads/compares a
large 10Mbyte file to and from the server. A full description is attached.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Set two Win2K clients systems up to connect to the Linux share over SMB.
2. Run a DOS application that repeatably writes a large file (10MBytes) to the
share and reads the file back comparing the file with the original.
3. On both of the Windows clients, explore the share with Explorer, and
repeatable hit F5 to refresh the share contents while the DOS application is
Actual Results: After a couple of minutes, one or both of the clients will
report a delayed-write error. This event will be logged in the event logger.
Also, the server Samba log file for the client will report a read_data error.
See attached full description for more information.
Expected Results: Clients should not experience write errors.
Created attachment 45619 [details]
Continued testing with non-Linux server in same environment:
Windows 2000: no client errors
OpenUnix8 w/ Samba 2.2.3a: no client errors.
Please test the 2.2.3a rpms available from http://people.redhat.com/teg/samba/
I see that 2.2.3a is mentioned in the "detailed description", but not mentioned
otherwise - did it help?
We tried 2.2.3a from samba.org on RH 7.2 and we get the same failure. Can we
assume the tha RH 2.2.3a rpm is the same as what's available on samba.org?
There shouldn't be code differences, no. Compiler flags and version, perhaps,
but not more than that.
Continued testing w/ Linux servers in same environment:
RH 7.1 w/ Samba 2.0.7-36: passed, no client errors
RH 7.1 w/ Samba 2.2.2 (from samba.org): failed w/ usual client errors
Continued testing w/ different Samba versions:
This problem seems to have been introduced between 2.0.9 and 2.2.0.
Can you recheck with current Samba 2.2 CVS (pserver.samba.org)? If this fails,
can you raise this on email@example.com - it would be particuarly useful
if you can post a link to the test application.
Intel, Can you please attach copies of your testing scripts to this bugzilla, or
post a link to where they can be downloaded from?
I am experiencing the same issue using RH7.2 / Samba 2.2.1a upgraded to 2.2.3a
on a Dell 1550. The RH box is being accessed by W2K IIS 5.0 web servers (I
know). The only difference is that instead of a single large file, the IIS
servers are constantly accessing many (upwards of 40,000 per dir) smaller files
ranging from 10k to about 2Mb (all PDF's). Have there been any updates or
fixes to this bug as of yet?
We are seeing similar failure running X-copy test from Windows clients, the
server is IA64 box with Redhat 7.2 (64-bit).
The x-copy test is a simple batch file copies and comapre files from mounted
drive (Linux box) using samba to local drives (Windows box).
Thanks to some great help from the Samba folks, the original issue sighted
appears to have been fixed with the latest Samba. We verified the clients do
not experience errors using the tip-of-tree from samba.org (pull on SAMBA_2_2
and HEAD branches).
Jeremy @ samba.org provided the fix with this explaination: "I fixed an issue
where an ill-timed signal could cause a socket read to return with -1 and
errno = EINTR".
Would the other two submitters (csherwood, maribasavaih) please verify that
this latest samba addresses their issues as well, and update this Bug so we
can get closure on this issue. Thanks.
We have been running on v2.2.5pre since it came out (about 4-6 weeks), and we
have not seen this problem come back...looks to be fixed as far as I can tell.
samba-2.2.5-6 is currently in Rawhide.