kern/109762: [hang] deadlock in g_down -> ahd_action -> contigmalloc

Frank de Bot ppi at searchy.net
Thu Aug 23 13:10:12 PDT 2007


The following reply was made to PR kern/109762; it has been noted by GNATS.

From: Frank de Bot <ppi at searchy.net>
To: bug-followup at FreeBSD.org,  andrew+pr2 at supernews.net
Cc:  
Subject: Re: kern/109762: [hang] deadlock in g_down -&gt; ahd_action -&gt;
 contigmalloc
Date: Thu, 23 Aug 2007 21:34:51 +0200

 I think I've stumbled on the same bug.
 
 I've managed to reproduce this problem. Although I have not been able to 
 extract a dump.
 
 Basic hardware information: 2x2.8ghz Prestiona Xeon, 2x256Mb DDR REG ECC 
 Ram, IO-Controller: Supermicro AOC-SAT2-MV8
 
 The System runs FreeBSD-6.2, compiled with most recent 
 FreeBSD-6.2-STABLE sources, world as well as kernel. Kernel is GENERIC 
 with SMP option enabled in it.
 
 The way I reproduce it:
 
 With gvinum a raid5 is created with the following definition:
 
 .--
 drive drive1 device /dev/ad6
 drive drive2 device /dev/ad8
 drive drive3 device /dev/ad10
 drive drive4 device /dev/ad12
 drive drive5 device /dev/ad14
 drive drive6 device /dev/ad16
 drive drive7 device /dev/ad18
 volume raid5_vol
          plex org raid5 512k
          sd length 476939m drive drive1
          sd length 476939m drive drive2
          sd length 476939m drive drive3
          sd length 476939m drive drive4
          sd length 476939m drive drive5
          sd length 476939m drive drive6
          sd length 476939m drive drive7
 .--
 
 Ran the command newfs /dev/gvinum/raid5_vol  , no options selected.
 After this I've ran iozone, with the following parameters:
 
 iozone -r 8 -r 64 -r 4096 -r 16384 -i 0 -i 1 -i 2 -Ra -g 4G > iozonerun.txt
 
 Ran in the newly created FS. I've ran iozone 3 times with the output 
 redirected to a file on the same FS, 3 times the system went rebooting. 
 The last time I tried to enable the creation of a kernel dump. The 
 following screen appeared:
 
 .--
 Fatal trap 12: page fault while in kernel mod
 cpuid = 0; apic id = 00
 fault virtual address  = 0x0
 fault code             = supervisor write, page not present
 instruction point      = 0x28:0x58089e8
 stack pointer          = 0x28:0xd44c2cdc
 frame pointer          = 0x28:0xd44c2ce8
 code segment           = base 0x0, limit 0xfffff, type 0x1b
                         = DPL 0, pres 1, def32 1, gran 1
 processor eflags       = interrupt enabled, resume, IOPL = 0
 current process        = 4 (g_down)
 trap number            = 12
 panic: page fault
 cpuid = 0
 Uptime: 21h9m55s
 Dumper 511 MB (2 chunks)
 swap_pager: indefinite wait in buffer: bufobj: 0, blkno: 62, size: 4096
 swap_pager: indefinite wait in buffer: bufobj: 0, blkno: 62, size: 4096
 swap_pager: indefinite wait in buffer: bufobj: 0, blkno: 62, size: 4096
 swap_pager: indefinite wait in buffer: bufobj: 0, blkno: 62, size: 4096
 swap_pager: indefinite wait in buffer: bufobj: 0, blkno: 62, size: 4096
 .--
 
 Last lime continued until system was turned off manually
 
 Between the 3 runs I had a run where I redirected the output to the 
 root-FS. The crash DID NOT occured.
 
 
 
 Regards,
 
 Frank de Bot


More information about the freebsd-bugs mailing list