Re: kernel panic from vinum during 'restore'

2002-10-06 Thread fergus

i have a crash dump for this now if anyone is interested.  it's not
exactly the same trace as it seems to originate from a VOP_LINK request
this time but i guess it's the same problem.

On 06.10-16:02, n0g0013 wrote:
 On 05.10-22:16, n0g0013 wrote:
  don't currently have the kernel trace for this (i'm rebuilding to get
  one) and the dump's not making it to disk.  perhaps someone else could
  try and verify in the mean time.
 
 don't know if this is related to Mikhail Teterin's post a few hours ago
 but i have the ddb trace.  it may not be exactly accurate because i don't
 have a serial ddb and therefore have to copy the output from the screen.
 
 +++ ddb output from kernel panic +++
 panic: kmem_malloc(4096): kmem_map too small: 23134208 total allocated
 Debugger(panic)
 Stopped atDebugger+0q45:  xchgl   %ebx,in_Debugger.0
 db trace
 Debugger(c02cd335) at Debugger+0x45
 panic(c02e1402,1000,161,10012,1000) at panic+0x9f
 kmem_malloc(c0432078,1000,4,c49b47cc,c02759a8) at kmem_malloc+0xcc
 page_alloc(c05ac500,1000,c49b47bf,4,0) at page_alloc+0x1a
 slab_zalloc(c05ac500,4,c0daba00,c05ac5e8,c05ac500) at slab_zalloc+0x108
 uma_zalloc_internal(c05ac500,0,4,c0daba00) at uma_zalloc_internal+0x13c
 uma_zalloc_arg(c05ac500,0,4) at uma_zalloc_arg+0x2f4
 getnewvnode(c02e070e,c0f4c000,c0dc1100,c49b48b8,90) at getnewvnode+0x3e9
 ffs_vget(c0f4c000,486,2,c49b491c,1) at ffs_vget+0x71
 ffs_valloc(c1153378,8180,c0f98780,c49b491c) at ffs_valloc+0xed
 ufs_makeinode(8180,c1153378,c49b4bf8,c49b4c0c) at ufs_makeinode+0x5b
 ufs_create(c49b4a70,c49b4a90,c01df957,c49b4a70,c02f1260) at ufs_create+0x2b
 ufs_vnoperate(c49b3a70) at ufs_vnoperate+0x13
 VOP_CREATE(c1153378,c49b4bf8,c49b4c0c,c49b4ad4,229) at VOP_CREATE+0x37
 vn_open_cred(c49b4be4,c49b4ce4,180,c0f98780,c49b4cd0) at vn_open_cred+0x154
 vn_open(c49b4be4,c49b4ce4,180,c1bec3b0,c0de7b04) at vn_open+0x18
 kern_open(c05b0680,80ad0c1,0,602,1b6) at kern_open+0x18b
 open(c05b0680,c49b4d14,3,a83,292) at open+0x18
 syscall(2f,2f,2f,80ad0c1,0) at syscall+0x24a
 Xint0x80_syscall() at Xint0x80_syscall+0x1d
 --- syscall (5, FreeBSD ELF32, open), eip = 0x08053a43, esp = 0xbfbff73c, ebp = 
0xbfbff778 ---
 db
 --- ddb output from kernel panic ---
 
   * that's no fun *
 
 config is identical to before with a kernel rebuild from same sources
 ( some time last night ).  the error i was orginally seeing (over the
 last week or so) seemed to appear in the vinum_devstrategy but since
 the various vinum changes it appears to have migrated.
 
 -- 
 t
  t
  z

 Date: Sat, 5 Oct 2002 22:16:10 +0100
 Subject: kernel panic from vinum during 'restore'
 From: n0g0013 [EMAIL PROTECTED]
 To: current [EMAIL PROTECTED]
 
 got panic from vinum on a small striped drive with small dump/restores
 -- dump of usr fs to file and restore to vinum volume.  dump file is
 about 120m.
 
 kernel built from about -0230 sources but i've been seeing them since
 first switched current on about 2 weeks ago.
 
 don't currently have the kernel trace for this (i'm rebuilding to get
 one) and the dump's not making it to disk.  perhaps someone else could
 try and verify in the mean time.
 
 +++ kernel panic message +++
 panic: kmem_malloc(4096): kmem_map to small: 23179264 total allocated
 
 syncing disks... panic: allocbuf: buffer not busy
 Uptime: 18m12s
 Dumping 48 MB
 ata1: resetting devices ..
 panic: free locked buf
 Uptime: 18m12s
 Automatice Reboot in 15 seconds - press a key on the console to abort
 --- kernel panic message ---
 
 +++ vinum config +++
 # Vinum configuration of eyore.blahdeblah.demon.co.uk, saved at Sat Oct 5 22:02:24 
2002
 drive snub device /dev/ad0s1h
 drive junk device /dev/ad2s1h
 volume var
 volume usr
 volume home
 volume software
 plex name var.plex0 org striped 534s vol var
 plex name usr.plex0 org striped 534s vol usr
 plex name home.plex0 org striped 534s vol home
 plex name software.plex0 org striped 534s vol software
 sd name snub.s0 drive snub plex var.plex0 len 163404s driveoffset 265s plexoffset 0s
 sd name junk.s0 drive junk plex var.plex0 len 163404s driveoffset 265s plexoffset 
534s
 sd name snub.s1 drive snub plex usr.plex0 len 614100s driveoffset 164105s plexoffset 
0s
 sd name junk.s1 drive junk plex usr.plex0 len 614100s driveoffset 164105s plexoffset 
534s
 sd name snub.s2 drive snub plex home.plex0 len 409578s driveoffset 778505s 
plexoffset 0s
 sd name junk.s2 drive junk plex home.plex0 len 409578s driveoffset 778505s 
plexoffset 534s
 sd name snub.s3 drive snub plex software.plex0 len 1638312s driveoffset 1188083s 
plexoffset 0s
 sd name junk.s3 drive junk plex software.plex0 len 1638312s driveoffset 1188083s 
plexoffset 534s
 --- vinum config ---
 
 -- 
 t
  t
  z

-- 
t
 t
 z

To Unsubscribe: send mail to [EMAIL PROTECTED]
with unsubscribe freebsd-current in the body of the message



Re: [ GEOM tests ] disklabel warnings and vinum drives lost

2002-10-04 Thread fergus

On 04.10-14:20, Poul-Henning Kamp wrote:
 I suspect vinum uses this sysctl to get an inventory of disks in
 the system, so can I get you to try again making sure you have
 rev. 1.20 of src/sys/geom/geom_disk.c ?

still in the middle of the build but i don't think so -- it looks like
vinum is using BIO_READ with a fixed offset VINUM_LABEL_OFFSET to
retrieve a raw copy of the disklabel to parse.

i don't have any references but from what you've mentioned in commits
over the last week that's not really going to work.  if there are any
geom design outlines i can read (in a digest or news) then i may be
able to put together a patch.

i'll let you know the test results when i have some.

-- 
t
 t
 z

To Unsubscribe: send mail to [EMAIL PROTECTED]
with unsubscribe freebsd-current in the body of the message



testing point releases

2002-09-24 Thread Fergus Cameron

i'm pretty new to current so perhaps this in naive but are there any
point releases for testing --- i.e. releases that are known to build
properly? 

the reason i ask is that i've got a problem and am not a coder but
essentially have no way to know whether it is worth reporting; whether it
is fixed because i can't quickly get the latest buildable and then start
to bring specific components up to date and verify.

do you really need to be on top of it all to work off of current ? (i.e.
i should know where to go back to)

To Unsubscribe: send mail to [EMAIL PROTECTED]
with unsubscribe freebsd-current in the body of the message