my experience with FreeBSD Xen guest

[Adrian Chadd]

I've only just started seeing filesystem corruption under high load.
-current Xen survived buildworlds in my initial testing so I was
focusing more on fixing network related bugs.

It is quite possible there are bugs in the VM system and/or Xen block
device code which are causing instabilities.

Please file a FreeBSD PR so the problem isn't forgotten about.

Thanks,



Adrian

2009/6/17 Alex Urbanowicz <alex.urbanowicz at artegence.com>:
> Hello
>
> I've spent last four days trying to get a stable 8.0-CURRENT on a CentOS
> 5.3. I am not very versed in BSD, so some of the problems may come from
> applying generic Unix approach to BSD environment.
>
> * disk operations are unstable. I attach dmesg from boot of a freshly
> fscked filesystem (16G LVM on host), the two stack backtraces are
> routinely happening during boot sequence. Also massive operations like
> buildworld usually lead to more syscall errors and routine filesystem
> corruption
>
> * sometimes the guest just crashes to kernel debugger, I have not
> isolated any particular reason why it happens
>
> * during the four days I only managed to successfully compile kernel
> once and world twice, installworld always fails, the last one with
> following error:
>
> [...]
> ===> secure (install)
> ===> secure/lib (install)
> ===> secure/lib/libcrypto (install)
> ".depend", line 19879: Unassociated shell command "��ɫ�������'�F/^�>�
> 闆 �� �
> ".depend", line 19880: Need an operator
> ".depend", line 19881: Need an operator
> ".depend", line 19883: Missing dependency operator
> ".depend", line 19884: Need an operator
> No closing parenthesis in archive specification
> ".depend", line 19885: Error in archive specification: " %
> ����Z��F��� ��_o���� o�x;����w,� �Fގ�/ 7�͚��"
> No closing parenthesis in archive specification
> ".depend", line 19886: Error in archive specification: "�����"
> ".depend", line 19888: Need an operator
> ".depend", line 19889: Need an operator
> ".depend", line 19893: Need an operator
> ".depend", line 19894: Missing dependency operator
> No closing parenthesis in archive specification
> ".depend", line 19899: Error in archive specification:
> "6pH�EX�� ��]����9� 䝺8�"
> ".depend", line 19910: Need an operator
> make: fatal errors encountered -- cannot continue
> *** Error code 1
>
> Stop in /usr/src/secure/lib.
> *** Error code 1
>
> Stop in /usr/src/secure.
> *** Error code 1
>
> Stop in /usr/src.
> *** Error code 1
>
> Stop in /usr/src.
> *** Error code 1
>
> Stop in /usr/src.
> *** Error code 1
>
> Stop in /usr/src.
>
> Also, during the mentioned buildworld the following syscall error
> happened:
>
> lock order reversal:
>  1st 0xc393b81c snaplk (snaplk)
> @ /usr/home/alex/bsd/sys/kern/vfs_vnops.c:295
>  2nd 0xc441037c ufs (ufs)
> @ /usr/home/alex/bsd/sys/ufs/ffs/ffs_snapshot.c:1587
> KDB: stack backtrace:
> X_db_sym_numargs(c0356957,e4d2f640,c0116b95,c01072bb,c035979e,...) at
> X_db_sym_numargs+0x146
> kdb_backtrace(c01072bb,c035979e,c36ccf10,c36ccbd0,e4d2f69c,...) at
> kdb_backtrace+0x29
> witness_display_spinlock(c035979e,c441037c,c034c8dc,c36ccbd0,c0371652,...) at witness_display_spinlock+0x75
> witness_checkorder(c441037c,9,c0371652,633,0,...) at witness_checkorder
> +0x839
> __lockmgr_args(c441037c,80000,0,0,0,...) at __lockmgr_args+0x797
> ffs_snapremove(c4410324,0,e4d2f7d8,e4d2f7d8,c035e67e,...) at
> ffs_snapremove+0x11f
> ffs_truncate(c4410324,0,0,c00,0,...) at ffs_truncate+0x5f7
> ufs_inactive(e4d2fa58,c4410398,c4410324,c4410398,e4d2fa70,...) at
> ufs_inactive+0x218
> VOP_INACTIVE_APV(c03b0200,e4d2fa58,c0360aae,925,c03c2160,...) at
> VOP_INACTIVE_APV+0xa5
> vref(c03b0200,e4d2fa8c,c0360aae,8ab,127,...) at vref+0x136e
> vput(c4410324,e4d2fac8,c0361c78,127,0,...) at vput+0x1cd
> vn_close(c4410324,1,c3b02600,c3b1f240,c0c74880,...) at vn_close+0x19a
> vn_close(c39611f8,c3b1f240,3,0,c39611f8,...) at vn_close+0x294
> _fdrop(c39611f8,c3b1f240,e4d2fb94,c01169dc,0,c0507a20,c3b1f2e4,c0507a20,c039b680,c4bd432c,6f4,c034e229,e4d2fbbc,c00de690,c4bd432c,8,c034e229,6f4) at _fdrop+0x43
> closef(c39611f8,c3b1f240,6f4,6f1,c3b1f2e4,...) at closef+0x290
> fdfree(c3b1f240,0,c034eafc,103,1,...) at fdfree+0x3ea
> exit1(c3b1f240,0,e4d2fd3c,c0329e23,c3b1f240,...) at exit1+0x533
> sys_exit(c3b1f240,e4d2fd08,4,c035a0f3,c039227c,...) at sys_exit+0x1d
> syscall(e4d2fd48) at syscall+0x2a3
> Xint0x80_syscall() at Xint0x80_syscall+0x22
> --- syscall (1, FreeBSD ELF32, sys_exit), eip = 0x28129f63, esp =
> 0xbf7fe33c, ebp = 0xbf7fe348 ---
>
> I guess that the both of the errors are related to fs corruption that
> happened during the course of buildworld (before I started buildworld I
> launched the guest from different image and fscked the / clean).
>
> The filesystem problems are severe, and block us from using FreeBSD/Xen.
>
> Regards
>
> Alex
>
> _______________________________________________
> freebsd-xen at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-xen
> To unsubscribe, send any mail to "freebsd-xen-unsubscribe at freebsd.org"
>
>