Package: libapache2-mod-fastcgi Version: 2.4.2-8 Severity: normal -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
I noticed that on our server fastcgi sometimes aborts because the select system call was interrupted instead of retrying the system call: [Fri Oct 31 20:14:47 2008] [error] [client xx.xxx.xx.xx] (4)Interrupted system call: FastCGI: comm with server "/xxx[...]/php5" aborted: select() failed, referer: http://www.aniki.info/Gott_Gauss I googled this and found the following patch: http://article.gmane.org/gmane.comp.web.fastcgi.devel/2514 However I think sleeping 1s before retrying is unnecessary (and having a maximum amount of retries might be a bit paranoid), so I'm not saying this patch should be applied as-is. I looked at the libapache-mod-fastcgi-2.4.6 source package and AFAICS it also has this bug. Note that the select(2) manpage clearly states: |Under Linux, select() may report a socket file descriptor as "ready for |reading", while nevertheless a subsequent read blocks. This could for |example happen when data has arrived but upon examination has wrong |checksum and is discarded. There may be other circumstances in which a |file descriptor is spuriously reported as ready. Thus it may be safer |to use O_NONBLOCK on sockets that should not block. So this applies even though the linked mail only talks about this problem occuring on AIX. FYI I'm not submitting the bug directly from the server (because I get 'out of memory' errors from reportbug due to a 32MB memory ulimit as user and I don't want to reportbug as root), the server is running etch/stable, I have edited the 'System Information' below accordingly. - -- System Information: Debian Release: etch/stable Architecture: i686 (x86) Kernel: Linux 2.6.25.4 Locale: LANG=en_IN, LC_CTYPE=en_IN (charmap=UTF-8) Shell: /bin/sh linked to /bin/bash -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.9 (GNU/Linux) iD8DBQFJC2xnzQZOfTz8JZwRAu+DAKCjdlP3YHXbAZ8J+odPMTiX5ShjHQCfZWzZ fxDZ/qLegIKOqCAORtott7s= =HwyU -----END PGP SIGNATURE----- -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]