Samba 3.0.27 daemon smb_panic

[Rcomian]

I have a server running samba and several linux based client machines (no windows ones).

When updating the server to samba 3.0.27, the clients can no-longer perform any operations on the shares (ls, etc), even though they mount fine.
The client pauses a long time then prints:

dragon var # ls network/
ls: reading directory network/: Input/output error

The server has this in the corresponding /var/long/samba/long.smbd:

[2007/11/16 21:28:00, 1] smbd/service.c:make_connection_snum(1033)
192.168.0.70 (192.168.0.70) connect to service network initially as user nobody (uid=65534, gid=250) (pid 18200)
[2007/11/16 21:28:03, 0] lib/util.c:smb_panic(1632)
PANIC (pid 18200): push_ascii - dest_len == -1
[2007/11/16 21:28:03, 0] lib/util.c:log_stack_trace(1736)
BACKTRACE: 14 stack frames:
#0 /usr/sbin/smbd(log_stack_trace+0x2e) [0x802357ae]
#1 /usr/sbin/smbd(smb_panic+0x5e) [0x802358de]
#2 /usr/sbin/smbd [0x8021e5cc]
#3 /usr/sbin/smbd(push_string_fn+0x4c) [0x8021e61c]
#4 /usr/sbin/smbd(srvstr_push_fn+0x54) [0x800b7664]
#5 /usr/sbin/smbd [0x800a22d1]
#6 /usr/sbin/smbd [0x800a3c17]
#7 /usr/sbin/smbd(handle_trans2+0x1e0) [0x800a43e0]
#8 /usr/sbin/smbd(reply_trans2+0x6a9) [0x800aa709]
#9 /usr/sbin/smbd [0x800c6c82]
#10 /usr/sbin/smbd(smbd_process+0x881) [0x800c8031]
#11 /usr/sbin/smbd(main+0xbfc) [0x802fc94c]
#12 /lib/libc.so.6(__libc_start_main+0xdc) [0xb7bd9fdc]
#13 /usr/sbin/smbd [0x80051581]
[2007/11/16 21:28:03, 0] lib/fault.c:dump_core(181)
dumping core in /var/log/samba/cores/smbd
[2007/11/16 21:28:12, 1] smbd/service.c:make_connection_snum(1033)

Despite the claims of a core file, none exist.

Downgrading to net-fs/samba-3.0.24-r3 & net-fs/samba-3.0.26a-r2 works perfectly, it's just the .27 version that goes wrong (I've done a couple of up/down grade cycles and checked on the clients). I've checked that I'm restarting the service and remounting the shares during all the checks. I'm going to mask .27 for now, but I'd really like to get some idea as to where to start looking for issues.

Server is 2.6.22-gentoo-r9 #2 PREEMPT Fri Nov 16 21:01:07 GMT 2007 i686 AMD Athlon(tm) XP 2800+ AuthenticAMD GNU/Linux.

Thanks.

[neiljw]

I have a server running samba and several linux based client machines (no windows ones).

So why on Earth are you using samba?

[hACknAT]

I have same problems here, my windows machines can connect but my linux can't.

[DaNIsH]

Also having this problem. In my case using Samba because I have a networked media player and I'm too lazy to set up UPnP

[tkhobbes]

Same problem here. I found this while googling, maybe it's an upstream bug? However, I could not find anything on bugs.gentoo.org - so I opened one (bug 199450)... hope I did it ok, as I never used bugzilla before (I only cc-ed myself to some bugs...)

[bubbl07]

I encountered this problem as well as soon as I upgraded to samba-3.0.27, but only on my Gentoo desktop. It worked fine on the media players, Windows machines, and OS X machines on the network.

It seems to be caused when using the deprecated smbfs. I compiled cifs into my kernel (File systems --> Network File Systems --> CIFS support), emerged mount-cifs (although I'm not sure if that's necessary), changed the filesystem type of the samba mounts from "smbfs" to "cifs" in /etc/fstab, and mounted and everything seemed to work fine.

I'm using hardened-sources-2.6.22-r8 on the samba server and gentoo-sources-2.6.23-r2 on the client.

Now, I'm not sure if this is the proper way of resolving the issue, but it seems to be working. Suggestions or advice welcome =).

[Rcomian]

Yes, changing to CIFS works for me too. However the mount parameters aren't exactly the same, so I'll have to expend energy to move, so for now I'm masking the dodgy server version - I'm sure in most cases using CIFS is the way to go tho.

[tkhobbes]

If I use "cifs" instead of "smbfs" in /etc/fstab, I get this error:

Code: Select all

mount: wrong fs type, bad option, bad superblock on //server/homeshare,
       missing codepage or other error
       In some cases useful info is found in syslog - try
       dmesg | tail  or so

cifs is compiled into the kernel, though....

[Rcomian]

Yeah I got that at first as well, to fix it you'll need to modify your kernel config to include CIFS as well as SMBFS (in networking filesystems).
Save the config, make modules, make modules_install, then modprobe cifs.

It should at least work after that.

[tkhobbes]

I have both compiled directly into the kernel (not as modules):

Code: Select all

CONFIG_SMB_FS=y
# CONFIG_SMB_NLS_DEFAULT is not set
CONFIG_CIFS=y
# CONFIG_CIFS_STATS is not set
# CONFIG_CIFS_WEAK_PW_HASH is not set
# CONFIG_CIFS_XATTR is not set
# CONFIG_CIFS_DEBUG2 is not set
# CONFIG_CIFS_EXPERIMENTAL is not set

[Rcomian]

Ok, you got me, I've no idea why that wouldn't work.

[bubbl07]

If I use "cifs" instead of "smbfs" in /etc/fstab, I get this error:

Code: Select all

mount: wrong fs type, bad option, bad superblock on //server/homeshare,
       missing codepage or other error
       In some cases useful info is found in syslog - try
       dmesg | tail  or so

cifs is compiled into the kernel, though....

Did you emerge the net-fs/mount-cifs package?

[stoil]

Does this have something todo with downgrade of samba in portage from 3.0.27 to 3.0.26a-r2?

[Rcomian]

Stoil: Yes, this is exactly the reason for the downgrade.
See: http://bugs.gentoo.org/show_bug.cgi?id=199450

Several linux smbfs based clients don't work with .27, and migrating to cifs is not-necessarily a no-brainer.
I believe the main reason for the .27 release was to fix a security issue. 26-r2 has this security patch in as well, but just that patch.

Upstream has a patch to fix .27, I'll not get a chance to look at it until tomorrow night.