Re: kmem_malloc(16384): kmem_map too small

On Sat, Nov 29, 2008 at 03:45:06PM +0100, Ivan Voras wrote:
> Lars Engels wrote:
> > With a ~3 week old current I got the following panic while running qemu:
> > 
> > panic: kmem_malloc(16384): kmem_map too small: 335536128 total allocated
> 
> > #11 0xc05d5b16 in panic (fmt=0xc090980a "kmem_malloc(%ld): kmem_map too small: %ld total allocated") at /usr/src/sys/kern/kern_shutdown.c:559
> > #12 0xc0802f9a in kmem_malloc (map=0xc1490084, size=16384, flags=1026) at /usr/src/sys/vm/vm_kern.c:303
> > #13 0xc07f9c37 in page_alloc (zone=0x0, bytes=16384, pflag=0xe53a09d7 "\002", wait=1026) at /usr/src/sys/vm/uma_core.c:952
> > #14 0xc07fc720 in uma_large_malloc (size=16384, wait=1026) at /usr/src/sys/vm/uma_core.c:2706
> > #15 0xc05c4d08 in malloc (size=16384, mtp=0xc0955f40, flags=1026) at /usr/src/sys/kern/kern_malloc.c:393
> > #16 0xc07db265 in softdep_disk_io_initiation (bp=0xd8228210) at /usr/src/sys/ufs/ffs/ffs_softdep.c:3815
> > #17 0xc07dfebc in ffs_geom_strategy (bo=0xc461a3cc, bp=0xd8228210) at buf.h:404
> > #18 0xc07efdd3 in ufs_strategy (ap=0xe53a0b90) at /usr/src/sys/ufs/ufs/ufs_vnops.c:2027
> > #19 0xc088f12d in VOP_STRATEGY_APV (vop=0xc0957320, a=0xe53a0b90) at vnode_if.c:1771
> > #20 0xc063f50e in bufstrategy (bo=0xc6259b20, bp=0xd8228210) at vnode_if.h:920
> > #21 0xc06456e1 in bufwrite (bp=0xd8228210) at buf.h:397
> > #22 0xc063ea48 in bawrite (bp=0xd8228210) at buf.h:385
> > #23 0xc07e4d6c in ffs_syncvnode (vp=0xc6259a78, waitfor=1) at /usr/src/sys/ufs/ffs/ffs_vnops.c:264
> > #24 0xc07e4f7c in ffs_fsync (ap=0xe53a0c5c) at /usr/src/sys/ufs/ffs/ffs_vnops.c:185
> > #25 0xc088e312 in VOP_FSYNC_APV (vop=0xc0956e00, a=0xe53a0c5c) at vnode_if.c:1007
> > #26 0xc0662aa9 in fsync (td=0xc46a28c0, uap=0xe53a0cf8) at vnode_if.h:529
> > #27 0xc0881555 in syscall (frame=0xe53a0d38) at /usr/src/sys/i386/i386/trap.c:1076
> > #28 0xc0866d60 in Xint0x80_syscall () at /usr/src/sys/i386/i386/exception.s:261
> > #29 0x00000033 in ?? ()
> > Previous frame inner to this frame (corrupt stack?)
> > 
> > # uname -a
> > FreeBSD maggie.bsd-geek.de 8.0-CURRENT FreeBSD 8.0-CURRENT #2: Tue Nov  4 22:52:12 CET 2008     lars_at_maggie.bsd-geek.de:/usr/obj/usr/src/sys/MAGGIE  i386
> 
> It looks like qemu for some reason causes your system to allocate a lot
> of memory for kernel's internal operation (either that or there's a
> memory leak somewhere). You should increase kmem_size just as you would
> do for the same situation with ZFS (see
> http://wiki.freebsd.org/ZFSTuningGuide).
> 
> Just for information: do you use kqemu?

Thanks, I set kmem_size to 512M now and will tell if the panic happens
again.
Yes, I use kqemu: kqemu-kmod-1.3.0.p11_9