Bug 104733
| Summary: | Most programs hang on DNS lookups | ||
|---|---|---|---|
| Product: | [Retired] Red Hat Raw Hide | Reporter: | Tom Lofgren <rh-bugs> |
| Component: | glibc | Assignee: | Jakub Jelinek <jakub> |
| Status: | CLOSED WORKSFORME | QA Contact: | Brian Brock <bbrock> |
| Severity: | medium | Docs Contact: | |
| Priority: | medium | ||
| Version: | 1.0 | CC: | fweimer |
| Target Milestone: | --- | ||
| Target Release: | --- | ||
| Hardware: | i686 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Fixed In Version: | Doc Type: | Bug Fix | |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2004-09-28 06:12:23 UTC | Type: | --- |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
I forgot to mention that just like in the case of the issue mentioned on the linux kernel list (linked), if I hit Ctrl-Z and either fg or bg the job after that, it temporarily unhangs. dig, for example, then actually returns the ip addresses, and then hangs again (where one would expect it to exit. Mozilla loads the page, and then hangs. Subsequent attempts of doing the same thing seem ineffective. There have been several bugs fixed in the thread library. I cannot say whether you'll find your problems are gone since I never had such problems myself and nobody else reported anything like this. And the fact that your strace shows the use of SIGHUP makes it unlikely that this is a libpthread problem: we do not use SIGHUP in libpthread. Anyway, try the code in FC3t2 or maybe the last FC2 update. If there is still a problem, reopen the bug with some more details and determine where the process got stuck. As mentioned above, I do not think this call got stuck in libpthread or anywhere in libc. |
Description of problem: Lots of applications hang on DNS lookups. Most applications that do DNS queries are affected, such as dig, nslookup, mozilla, etc. Interestingly enough, the telnet and ssh command-line clients work fine. Version-Release number of selected component (if applicable): glibc-2.3.2-82.i686.rpm, but I believe I saw it with a slightly older rawhide versions also. The one in stock RH 9, glibc-2.3.2-11.9.i686.rpm is good. How reproducible: I can reproduce it every time on my machine. In fact, I can't get the programs to *not* hang Steps to Reproduce: 1. dig www.cnn.com Actual results: No output: process is hung. Expected results: Regular dig output of IP addresses. Additional info: Here is the last part of an strace of the dig: open("/etc/resolv.conf", O_RDONLY) = 5 fstat64(5, {st_mode=S_IFREG|0644, st_size=105, ...}) = 0 mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x40016000 read(5, "; generated by /sbin/dhclient-sc"..., 4096) = 105 read(5, "", 4096) = 0 close(5) = 0 munmap(0x40016000, 4096) = 0 rt_sigaction(SIGHUP, {0x402111d0, ~[RTMIN], SA_RESTORER, 0x402479c8}, NULL, 8) = 0 rt_sigsuspend([] <unfinished ...> So it hangs in rt_sigsuspend(). When I first googled for this, I came across http://www.ussg.iu.edu/hypermail/linux/kernel/0308.0/0742.html This was what made me downgrade glibc and try again. Using the stock RH9, everything works fine. It seems to me that this is an old bug in the threading code that has been reintroduced, but I'll let you guys draw the conclusions. It also striked me as a bit odd that no one else has noticed this, since a commonly used app like mozilla won't run. Maybe it's something that's specific to my configuration?