lkml.org 
[lkml]   [2000]   [Sep]   [22]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
Subject[2.2.17] poll(2) bug/race condition?
Playing with Uwe Ohse's ftpcopy 0.3.2 (which I like a lot), I found
strange behaviour in poll(). Does not happen under FreeBSD 4.0.

Preface: ftpcopy is an FTP client (mirroring agent, actually) that uses
poll(). Available on http://www.ohse.de/uwe/ftpcopy.html

Now, if I suspend ftpcopy (^Z in bash) and resume it like 5 seconds
later, I get ENOSYS back from poll. (sometimes, strace does not log the
signal, poll then returns EINTR properly.)

The problem in this is that strace-4.1 that I've been using is
documented that the process ignores SIGSTOP, but I don't currently know
of another way to trigger this behaviour.

Kernel 2.2.17 with openwall security patch, reiserfs, ide, lm_sensors,
i2c patches and dc390 driver 2.0e2.

WARNING: If you try to reproduce this, CALL FTPCOPY FROM AN EMPTY
-------- DIRECTORY! ftpcopy removes files that are not present on the
server, that's why, thus, if you run ftpcopy with `pwd` =
/usr/src/linux, it will hurt you bad.

Reding through Linux' and SUS v2's poll() man pages, ENOSYS is not
specified as error condition, just ENOMEM, EFAULT, EINTR for Linux. SUS
v2 documents EAGAIN, EINVAL and EINTR.

strace log is as follows:

poll([{fd=3, events=POLLIN, revents=POLLIN}], 1, 30020) = 1
read(3, "250 CWD command successful.\r\n", 8192) = 29
gettimeofday({969616706, 131134}, NULL) = 0
poll([{fd=3, events=POLLOUT, revents=POLLOUT}], 1, 30020) = 1
write(3, "LIST\r\n", 6) = 6
gettimeofday({969616706, 132726}, NULL) = 0
poll( <unfinished ...>
--- SIGCONT (Fortgesetzt) ---
<... poll resumed> [{fd=3, events=POLLIN}], 1, 30020) = -1 ENOSYS
(Function not implemented)
ftpcopywrite(2, ": ", 2: ) = 2
write(2, "read (LIST answer)", 18read (LIST answer)) = 18
write(2, ": ", 2: ) = 2
write(2, "timed out", 9timed out) = 9
write(2, "\n", 1
) = 1
_exit(1) = ?

--
Matthias Andree
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 12:38    [W:0.044 / U:3.816 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site