Seeing lots of coredumps from samba when using upstream cifs
Subsystems (2)
Hi Jeremy, Steve,
I've been trying to investigate all the reconnection issues cifs is seeing in the currently upstream kernel from running the generic/013 xfstest against it, and I've realised Samba is coredumping a lot (attached is one example, at lease several others look similar).
The version of the Fedora Samba RPM I'm using is: samba-4.21.4-1.fc41.x86_64
In smb2_lease_key_equal (frame #10), k1 is 0x18 - which is presumably the root of the SEGV.
Going up to frame #12 (delay_rename_lease_break_fn), fsp->lease is NULL.
(gdb) p *fsp
$5 = {next = 0x558d634a1520, prev = 0x558d6345f130, fnum = 566610366, op = 0x558d63537b20, conn = 0x558d63414680, fh = 0x558d635182c0, num_smb_operations = 0,
file_id = {devid = 48, inode = 1875968, extid = 0}, initial_allocation_size = 0, file_pid = 52138, vuid = 3019447861, open_time = {tv_sec = 1750343292,
tv_usec = 868402}, access_mask = 65536, fsp_flags = {is_pathref = true, is_fsa = true, have_proc_fds = true, kernel_share_modes_taken = false, update_write_time_triggered = false, update_write_time_on_close = false, write_time_forced = false, can_lock = true, can_read = false, can_write = false, modified = false, is_directory = false, is_dirfsp = false, aio_write_behind = false, initial_delete_on_close = false, delete_on_close = false, is_sparse = true, backup_intent = false, use_ofd_locks = true, closing = false, lock_failure_seen = false, encryption_required = false, fstat_before_close = false}, update_write_time_event = 0x0, close_write_time = {tv_sec = 0, tv_nsec = 1073741822}, oplock_type = 0, leases_db_seqnum = 0,
lease_type = 0, lease = 0x0, sent_oplock_break = 0, oplock_timeout = 0x0, current_lock_count = 0, posix_flags = 0, fsp_name = 0x558d63566550, name_hash = 3474280865, mid = 1043, vfs_extension = 0x0, fake_file_handle = 0x0, notify = 0x0, base_fsp = 0x0, stream_fsp = 0x0, share_mode_flags_seqnum = 0, share_mode_flags = 0, brlock_seqnum = 0, brlock_rec = 0x558d63518220, dptr = 0x0, print_file = 0x0, num_aio_requests = 0, aio_requests = 0x0, blocked_smb1_lock_reqs = 0x0, lock_failure_offset = 0}
David
---
#0 __pthread_kill_implementation (threadid=<optimized out>, signo=signo@entry=6, no_tid=no_tid@entry=0) at pthread_kill.c:44
#1 0x00007f44435dcbc3 in __pthread_kill_internal (threadid=<optimized out>, signo=6) at pthread_kill.c:78
#2 0x00007f4443583f9e in __GI_raise (sig=sig@entry=6) at ../sysdeps/posix/raise.c:26
#3 0x00007f444356b942 in __GI_abort () at abort.c:79
#4 0x00007f4443ac26dc in dump_core () at ../../source3/lib/dumpcore.c:339
#5 0x00007f4443ac2734 in smb_panic_s3 (why=<optimized out>) at ../../source3/lib/util.c:716
#6 0x00007f44437c56b6 in smb_panic (why=why@entry=0x7ffcde469310 "Signal 11: Segmentation fault") at ../../lib/util/fault.c:209
#7 0x00007f44437c575d in fault_report (sig=11) at ../../lib/util/fault.c:83
#8 sig_fault (sig=11) at ../../lib/util/fault.c:94
#9 <signal handler called>
#10 0x00007f44429acbf9 in smb2_lease_key_equal (k1=k1@entry=0x18, k2=k2@entry=0x7ffcde469b98) at ../../libcli/smb/smb2_lease.c:93
#11 0x00007f44429acc47 in smb2_lease_equal (g1=<optimized out>, k1=k1@entry=0x18, g2=g2@entry=0x7ffcde469b84, k2=k2@entry=0x7ffcde469b98)
at ../../libcli/smb/smb2_lease.c:101#12 0x00007f4443c61c5f in delay_rename_lease_break_fn (e=0x7ffcde469b60, private_data=0x7ffcde469cb0) at ../../source3/smbd/smb2_setinfo.c:202 #13 0x00007f4443cdcb8f in share_mode_for_one_entry (fn=<optimized out>, private_data=<optimized out>, i=<synthetic pointer>, data=0x558d635920ad "\025\262\021",
num_share_modes=<synthetic pointer>, writeback=<synthetic pointer>) at ../../source3/locking/share_mode_lock.c:2161
#14 share_mode_forall_entries (lck=<optimized out>, fn=0x7f4443ccdf40 <share_mode_forall_leases_fn>, private_data=0x7ffcde469c30)
at ../../source3/locking/share_mode_lock.c:2265
#15 0x00007f4443cd4556 in share_mode_forall_leases (lck=0x558d635b2b80, fn=0x7f4443c61c20 <delay_rename_lease_break_fn>, private_data=0x7ffcde469cb0)
at ../../source3/locking/locking.c:1349
#16 0x00007f4443c61d42 in delay_rename_for_lease_break (req=req@entry=0x558d635b2850, smb2req=smb2req@entry=0x558d635b10b0, ev=ev@entry=0x558d63425d80,
fsp=fsp@entry=0x558d6357a960, lck=lck@entry=0x558d635b2b80, data=data@entry=0x558d6349da00 "\001", data_size=116) at ../../source3/smbd/smb2_setinfo.c:245#17 0x00007f4443c62721 in smbd_smb2_setinfo_send (mem_ctx=0x558d635b10b0, ev=0x558d63425d80, smb2req=0x558d635b10b0, fsp=0x558d6357a960, --Type <RET> for more, q to quit, c to continue without paging--
in_info_type=<optimized out>, in_file_info_class=<optimized out>, in_input_buffer=..., in_additional_information=<optimized out>) at ../../source3/smbd/smb2_setinfo.c:491
#18 smbd_smb2_request_process_setinfo (req=0x558d635b10b0) at ../../source3/smbd/smb2_setinfo.c:112
#19 0x00007f4443c476d4 in smbd_smb2_request_dispatch (req=0x558d635b10b0) at ../../source3/smbd/smb2_server.c:3582 #20 0x00007f4443c48f99 in smbd_smb2_request_dispatch_immediate (ctx=ctx@entry=0x558d63425d80, im=<optimized out>, im@entry=0x558d635b2760,
private_data=private_data@entry=0x558d635b10b0) at ../../source3/smbd/smb2_server.c:3910
#21 0x00007f4443769a50 in tevent_common_invoke_immediate_handler (im=0x558d635b2760, removed=removed@entry=0x0) at ../../tevent_immediate.c:190 #22 0x00007f4443769ab2 in tevent_common_loop_immediate (ev=ev@entry=0x558d63425d80) at ../../tevent_immediate.c:236
#23 0x00007f444376d750 in epoll_event_loop_once (ev=0x558d63425d80, location=<optimized out>) at ../../tevent_epoll.c:905 #24 0x00007f44437648e4 in std_event_loop_once (ev=0x558d63425d80, location=0x7f4443bae1c8 "../../source3/smbd/smb2_process.c:2163")
at ../../tevent_standard.c:110
#25 0x00007f4443766499 in _tevent_loop_once (ev=ev@entry=0x558d63425d80, location=location@entry=0x7f4443bae1c8 "../../source3/smbd/smb2_process.c:2163")
at ../../tevent.c:820
#26 0x00007f44437665cb in tevent_common_loop_wait (ev=0x558d63425d80, location=0x7f4443bae1c8 "../../source3/smbd/smb2_process.c:2163") at ../../tevent.c:949
#27 0x00007f4443764964 in std_event_loop_wait (ev=0x558d63425d80, location=0x7f4443bae1c8 "../../source3/smbd/smb2_process.c:2163")
at ../../tevent_standard.c:141
#28 0x00007f4443c38fa3 in smbd_process (ev_ctx=ev_ctx@entry=0x558d63425d80, msg_ctx=msg_ctx@entry=0x558d63420a50, sock_fd=sock_fd@entry=35,
interactive=interactive@entry=false) at ../../source3/smbd/smb2_process.c:2163
#29 0x0000558d47185606 in smbd_accept_connection (ev=0x558d63425d80, fde=<optimized out>, flags=<optimized out>, private_data=<optimized out>)
at ../../source3/smbd/server.c:1033#30 0x00007f4443769818 in tevent_common_invoke_fd_handler (fde=0x558d63439fe0, flags=1, removed=removed@entry=0x0) at ../../tevent_fd.c:174 #31 0x00007f444376da56 in epoll_event_loop (epoll_ev=0x558d63425fb0, tvalp=0x7ffcde46a380) at ../../tevent_epoll.c:696
#32 epoll_event_loop_once (ev=<optimized out>, location=<optimized out>) at ../../tevent_epoll.c:926
--Type <RET> for more, q to quit, c to continue without paging--
#33 0x00007f44437648e4 in std_event_loop_once (ev=0x558d63425d80, location=0x558d4717fbd0 "../../source3/smbd/server.c:1382") at ../../tevent_standard.c:110 #34 0x00007f4443766499 in _tevent_loop_once (ev=ev@entry=0x558d63425d80, location=location@entry=0x558d4717fbd0 "../../source3/smbd/server.c:1382")
at ../../tevent.c:820
#35 0x00007f44437665cb in tevent_common_loop_wait (ev=0x558d63425d80, location=0x558d4717fbd0 "../../source3/smbd/server.c:1382") at ../../tevent.c:949 #36 0x00007f4443764964 in std_event_loop_wait (ev=0x558d63425d80, location=0x558d4717fbd0 "../../source3/smbd/server.c:1382") at ../../tevent_standard.c:141 #37 0x0000558d47188acf in smbd_parent_loop (parent=0x558d6340e270, ev_ctx=0x558d63425d80) at ../../source3/smbd/server.c:1382
#38 main (argc=<optimized out>, argv=<optimized out>) at ../../source3/smbd/server.c:2354
Quoted 8 lines · David Howells
Hi Jeremy, Steve,
I've been trying to investigate all the reconnection issues cifs is seeing in the currently upstream kernel from running the generic/013 xfstest against it, and I've realised Samba is coredumping a lot (attached is one example, at lease several others look similar).
The version of the Fedora Samba RPM I'm using is: samba-4.21.4-1.fc41.x86_64
The problem seems to be fixed in samba-4.22.2-1.fc42.x86_64.
David
Quoted 4 lines · David Howells via samba-technical
Quoted 2 lines · David Howells
Hi Jeremy, Steve,
Quoted 9 lines
Quoted 6 lines
I've been trying to investigate all the reconnection issues cifs is seeing in the currently upstream kernel from running the generic/013 xfstest against it, and I've realised Samba is coredumping a lot (attached is one example, at lease several others look similar).
The version of the Fedora Samba RPM I'm using is: samba-4.21.4-1.fc41.x86_64
The problem seems to be fixed in samba-4.22.2-1.fc42.x86_64.
It looks a bit like
https://bugzilla.samba.org/show_bug.cgi?id=15803
which was fixed in 4.22 by patches for
https://bugzilla.samba.org/show_bug.cgi?id=15608
Douglas