Getting these fairly frequently on a Fedora 7 box: Aug 12 09:01:00 saga smbd[20225]: [2007/08/12 09:01:00, 0] lib/fault.c:fault_report(41) Aug 12 09:01:00 saga smbd[20225]: ====================================================== ========= Aug 12 09:01:00 saga smbd[20225]: [2007/08/12 09:01:00, 0] lib/fault.c:fault_report(42) Aug 12 09:01:00 saga smbd[20225]: INTERNAL ERROR: Signal 11 in pid 20225 (3.0.25b-2.fc7) Aug 12 09:01:00 saga smbd[20225]: Please read the Trouble-Shooting section of the Samba3 -HOWTO Aug 12 09:01:00 saga smbd[20225]: [2007/08/12 09:01:00, 0] lib/fault.c:fault_report(44) Aug 12 09:01:00 saga smbd[20225]: Aug 12 09:01:00 saga smbd[20225]: From: http://www.samba.org/samba/docs/Samba3-HOWTO.pdf Aug 12 09:01:00 saga smbd[20225]: [2007/08/12 09:01:00, 0] lib/fault.c:fault_report(45) Aug 12 09:01:00 saga smbd[20225]: ====================================================== ========= Aug 12 09:01:00 saga smbd[20225]: [2007/08/12 09:01:00, 0] lib/util.c:smb_panic(1654) Aug 12 09:01:00 saga smbd[20225]: PANIC (pid 20225): internal error Aug 12 09:01:00 saga smbd[20225]: [2007/08/12 09:01:00, 0] lib/util.c:log_stack_trace(1758 ) Aug 12 09:01:00 saga smbd[20225]: BACKTRACE: 12 stack frames: Aug 12 09:01:00 saga smbd[20225]: #0 smbd(log_stack_trace+0x2d) [0x802439fd] Aug 12 09:01:00 saga smbd[20225]: #1 smbd(smb_panic+0x5d) [0x80243b2d] Aug 12 09:01:00 saga smbd[20225]: #2 smbd [0x8022e53a] Aug 12 09:01:00 saga smbd[20225]: #3 [0x12d420] Aug 12 09:01:00 saga smbd[20225]: #4 /lib/libc.so.6(memset+0x37) [0x3d86f7] Aug 12 09:01:00 saga smbd[20225]: #5 smbd(write_file+0x979) [0x80067169] Aug 12 09:01:00 saga smbd[20225]: #6 smbd(reply_write_and_X+0x571) [0x80086261] Aug 12 09:01:00 saga smbd[20225]: #7 smbd [0x800c54d0] Aug 12 09:01:00 saga smbd[20225]: #8 smbd(smbd_process+0x836) [0x800c6566] Aug 12 09:01:00 saga smbd[20225]: #9 smbd(main+0xbdd) [0x8032079d] Aug 12 09:01:00 saga smbd[20225]: #10 /lib/libc.so.6(__libc_start_main+0xe0) [0x37ef70] Aug 12 09:01:00 saga smbd[20225]: #11 smbd [0x8004a181] Aug 12 09:01:00 saga smbd[20225]: [2007/08/12 09:01:00, 0] lib/fault.c:dump_core(181) Aug 12 09:01:00 saga smbd[20225]: dumping core in /var/log/samba/cores/smbd Aug 12 09:01:00 saga smbd[20225]: I believe this is during Ghost writing a backup to the share. (gdb) bt #0 0x0012d402 in __kernel_vsyscall () #1 0x00391fa0 in raise () from /lib/libc.so.6 #2 0x003938b1 in abort () from /lib/libc.so.6 #3 0x8022dff9 in dump_core () at lib/fault.c:192 #4 0x80243be5 in smb_panic (why=0x8037ca8c "internal error") at lib/util.c:1670 #5 0x8022e53a in sig_fault (sig=11) at lib/fault.c:47 #6 <signal handler called> #7 0x003d86f7 in memset () from /lib/libc.so.6 #8 0x80067169 in write_file (fsp=0x80e43c70, data=0x80e53954 "", pos=113377280, n=4181983233) at /usr/include/bits/string3.h:96 #9 0x80086261 in reply_write_and_X (conn=0x80e941d0, inbuf=0x80e53910 "", outbuf=0x80e73d58 "", length=69, bufsize=131072) at smbd/reply.c:3135 #10 0x800c54d0 in switch_message (type=47, inbuf=0x80e53910 "", outbuf=0x80e73d58 "", size=69, bufsize=131072) at smbd/process.c:1003 #11 0x800c6566 in smbd_process () at smbd/process.c:1030 #12 0x8032079d in main (argc=-1109787819, argv=0xbfe3ff34) at smbd/server.c:1120 (gdb) print *fsp $2 = {next = 0x0, prev = 0x0, fnum = 5408, conn = 0x80e941d0, fh = 0x80debc38, num_smb_operations = 3390, rap_print_jobid = 0, dev = 2145, inode = 351842, initial_allocation_size = 0, mode = 33268, file_pid = 1628, vuid = 101, wbmpx_ptr = 0x0, wcp = 0x80e95080, open_time = {tv_sec = 1187114387, tv_usec = 808635}, access_mask = 131487, share_access = 1, pending_modtime_owner = 0, pending_modtime = {tv_sec = 0, tv_nsec = 0}, last_write_time = {tv_sec = 0, tv_nsec = 0}, oplock_type = 3, sent_oplock_break = 0, oplock_timeout = 0x0, last_lock_failure = {context = {smbpid = 0, tid = 0, pid = { pid = 0}}, start = 0, size = 0, fnum = 0, lock_type = READ_LOCK, lock_flav = WINDOWS_LOCK}, current_lock_count = 0, pending_break_messages = 0x0, num_pending_break_messages = 0, can_lock = 1, can_read = 1, can_write = 1, print_file = 0, modified = 1, is_directory = 0, is_stat = 0, aio_write_behind = 0, lockdb_clean = 0, initial_delete_on_close = 0, posix_open = 0, fsp_name = 0x80dee748 "nome/C_Drive118_i001.iv2i", vfs_extension = 0x0, fake_file_handle = 0x0, notify = 0x0}
Looks like the memsets in write_file are: memset(wcp->data + wcp->data_size, '\0', pos - (wcp->offset + wcp->data_size) ); and memset(wcp->data, '\0', wcp->data_size); (gdb) print *wcp $7 = {file_size = 112984064, offset = 112984064, alloc_size = 262144, data_size = 393216, data = 0x80eb5938 ""} (gdb) print pos $8 = 113377280
I'm assuming you have "write cache size = XXX" set in your smb.conf ? I recently fixed a bug in this (reported by another user). Can you check out the SAMBA_3_0_25 svn tree to test this ? I think this should be fixed for 3.0.25c but I'd like to be sure. You could also remove the write cache size parameter and check that also fixes it. Jeremy.
(In reply to comment #2) > I'm assuming you have "write cache size = XXX" set in your smb.conf ? Indeed. I've installed the latest SAMBA_3_0_25 svn branch and we'll give that a try. BTW - is it possible for a client to override the write cache setting? I'm trying to track down slow SMB performance from Mac OS X clients and it appears that their writes are written to disk in 4kb chunks despite the write cache size = 262144 setting for that share.
write cache is only used for oplocked files, and clients can defeat it be doing a scatter write pattern. I'd need to see debug level 10 logs to see why this is occurring. Jeremy.
Created attachment 2867 [details] smb log with log level = 10 This is a log from a OS X client doing a write of a 512k file. Looks like oplocks aren't being used at all, and hence no write cache. I don't know if there are any other things that would obviously impair performance. I only see about 3.4MB/s writing from a OS X client and around 8.8MB/s from a linux client.