David Brownlee | 2 Apr 16:38 2004

bonnie++ and sd0(esp0:0:3:0): esp0: timed out

	Running 2.0_BETA on a SPARCstation-20. Machine runs fine in service
	but wanted to check the disk would work fine under load.

	Booted with nfs root, made a 750M partition on the Seagate ST31200W
	and ran bonnie++. It consistently runs fine mounted normally, but
	when mounted with softdep it chokes with:

Using uid:0, gid:0.
Writing with putc()...done
Writing intelligently...sd0(esp0:0:3:0): esp0: timed out [ecb 0xf0a413b8 (flags 0x1, dleft 1000, stat
0)], <state 1, nexus 0x0, phase(l 10, c 100, p 7), resid 0, msg(q 0,o 0) >
sd0(esp0:0:3:0): esp0: timed out [ecb 0xf0a417a8 (flags 0x1, dleft 1000, stat 0)], <state 2, nexus
0xf0a41038, phase(l 0, c 100, p 3), resid 0, msg(q 0,o 100) DMA active>
sd0(esp0:0:3:0): esp0: timed out [ecb 0xf0a41ce8 (flags 0x1, dleft 1000, stat 0)], <state 1, nexus 0x0,
phase(l 10, c 100, p 7), resid 0, msg(q 0,o 0) >
sd0(esp0:0:3:0): esp0: timed out [ecb 0xf0a413f0 (flags 0x1, dleft 1000, stat 0)], <state 1, nexus 0x0,
phase(l 10, c 100, p 7), resid 0, msg(q 0,o 0) >
sd0(esp0:0:3:0): esp0: timed out [ecb 0xf0a415e8 (flags 0x1, dleft 1000, stat 0)], <state 2, nexus
0xf0a41770, phase(l 0, c 100, p 3), resid 0, msg(q 0,o 100) DMA active>
sd0(esp0:0:3:0): esp0: timed out [ecb 0xf0a41d20 (flags 0x1, dleft 1000, stat 0)], <state 1, nexus 0x0,
phase(l 10, c 100, p 7), resid 0, msg(q 0,o 0) >
sd0(esp0:0:3:0): esp0: timed out [ecb 0xf0a41428 (flags 0x1, dleft 1000, stat 0)], <state 1, nexus 0x0,
phase(l 10, c 100, p 7), resid 0, msg(q 0,o 0) >
sd0(esp0:0:3:0): esp0: timed out [ecb 0xf0a41690 (flags 0x1, dleft 1000, stat 0)], <state 2, nexus
0xf0a418c0, phase(l 0, c 100, p 3), resid 0, msg(q 0,o 100) DMA active>
sd0(esp0:0:3:0): esp0: timed out [ecb 0xf0a41a10 (flags 0x1, dleft 1000, stat 0)], <state 1, nexus 0x0,
phase(l 10, c 100, p 7), resid 0, msg(q 0,o 0) >
sd0(esp0:0:3:0): esp0: timed out [ecb 0xf0a411c0 (flags 0x101, dleft 0, stat 0)], <state 1, nexus 0x0,
phase(l 10, c 100, p 3), resid 0, msg(q 0,o 0) >

(Continue reading)

David Brownlee | 2 Apr 19:31 2004
Picon

Re: bonnie++ and sd0(esp0:0:3:0): esp0: timed out

	Further to this. Problem does not occur with async either.
	Does anyone have any thoughts? I maybe be able to try on an esp
	equipped ultra1 to see if it has similar issues.

On Fri, 2 Apr 2004, David Brownlee wrote:

> 	Running 2.0_BETA on a SPARCstation-20. Machine runs fine in service
> 	but wanted to check the disk would work fine under load.
>
> 	Booted with nfs root, made a 750M partition on the Seagate ST31200W
> 	and ran bonnie++. It consistently runs fine mounted normally, but
> 	when mounted with softdep it chokes with:
>
> Using uid:0, gid:0.
> Writing with putc()...done
> Writing intelligently...sd0(esp0:0:3:0): esp0: timed out [ecb 0xf0a413b8 (flags 0x1, dleft 1000,
stat 0)], <state 1, nexus 0x0, phase(l 10, c 100, p 7), resid 0, msg(q 0,o 0) >
> sd0(esp0:0:3:0): esp0: timed out [ecb 0xf0a417a8 (flags 0x1, dleft 1000, stat 0)], <state 2, nexus
0xf0a41038, phase(l 0, c 100, p 3), resid 0, msg(q 0,o 100) DMA active>
> sd0(esp0:0:3:0): esp0: timed out [ecb 0xf0a41ce8 (flags 0x1, dleft 1000, stat 0)], <state 1, nexus 0x0,
phase(l 10, c 100, p 7), resid 0, msg(q 0,o 0) >
> sd0(esp0:0:3:0): esp0: timed out [ecb 0xf0a413f0 (flags 0x1, dleft 1000, stat 0)], <state 1, nexus 0x0,
phase(l 10, c 100, p 7), resid 0, msg(q 0,o 0) >
> sd0(esp0:0:3:0): esp0: timed out [ecb 0xf0a415e8 (flags 0x1, dleft 1000, stat 0)], <state 2, nexus
0xf0a41770, phase(l 0, c 100, p 3), resid 0, msg(q 0,o 100) DMA active>
> sd0(esp0:0:3:0): esp0: timed out [ecb 0xf0a41d20 (flags 0x1, dleft 1000, stat 0)], <state 1, nexus 0x0,
phase(l 10, c 100, p 7), resid 0, msg(q 0,o 0) >
> sd0(esp0:0:3:0): esp0: timed out [ecb 0xf0a41428 (flags 0x1, dleft 1000, stat 0)], <state 1, nexus 0x0,
phase(l 10, c 100, p 7), resid 0, msg(q 0,o 0) >
> sd0(esp0:0:3:0): esp0: timed out [ecb 0xf0a41690 (flags 0x1, dleft 1000, stat 0)], <state 2, nexus
(Continue reading)

David Brownlee | 2 Apr 19:43 2004
Picon

Re: install on 8MB boxes

On Sun, 28 Mar 2004, Manuel Bouyer wrote:

> I've been able to install current from netboot on a 8MB sparc IPX, the only
> requirement is to use a smaller kernel than GENERIC or GENERIC_SCSI3.
> GENERIC doesn't load, it fails with:
> Booting netbsd
> Cannot load netbsd: error=27
> (27 is EFBIG)
>
> I don't think the problem is the amount of RAM, but it's probably the
> fact that it's 2 segments of 4MB with a hole between them.
> The kernel I used is just GENERIC with sun4m-specific things commented out,
> no RASTERCONSOLE, no KTRACE/SYSTRACE, no pseudo-filesystems (they can be
> loaded as LKM), and no network protocols others than INET and INET6.
> There may be a few more things that could be removed, like sbus cards known
> to work only on sun4m systems. This kernel should work for all sun4 and sun4c
> systems, and is 2762499 bytes (vs 3472644).
>
> Is it worth adding to the list of standard kernels and distribution,
> or do we assume peoples with low memory machines will be able to cross-compile
> their own kernel ?

	I do not think it is fair to force a new user to cross-compile
	just to test out NetBSD.

	Are there any sun4m boxes with less than 8M contiguous memory?
	The lowest spec sun4m would probably be the classic which requires
	at a minimum a pair of 4MB simms.

	Maybe have a SUN4C_SMALL with your changes and compiled with -Os?
(Continue reading)

David Brownlee | 2 Apr 19:44 2004
Picon

Re: SunOS4 libdl.so

On Fri, 19 Mar 2004, der Mouse wrote:

> > ftp://sunsolve.Sun.COM/pub/patches offers 102545-13.tar.Z and
> > 100257-06.tar.Z which contain nearly everything needed to run SunOS4
> > executables under NetBSD/sparc, [...]
>
> > While this may be of historic interest it would be nice to drop in a
> > pkgsrc entry for compat_sunos :)
>
> Did you check the /README as mentioned in their FTP splash screen?  The
> license under which those things travel does not permit using them
> except "with the specific software previously licensed by you from Sun
> with which this Modification is intended to operate".  To my eye this
> rules out using them with NetBSD/sparc SunOS compatability mode.
>
> Now, as an individual you may choose to ignore that and probably
> (probably!) will not get in much trouble for it.  But I think it would
> set a really bad precedent for NetBSD's pkgsrc to do so.

	Hmm - It _would_ need to have an appropriate licence in
	pkgsrc/licenses :)

--

-- 
		David/absolute          -- www.netbsd.org: No hype required --

Andrey Petrov | 2 Apr 21:17 2004
Picon

Re: bonnie++ and sd0(esp0:0:3:0): esp0: timed out

On Fri, Apr 02, 2004 at 06:31:55PM +0100, David Brownlee wrote:
> 	Further to this. Problem does not occur with async either.
> 	Does anyone have any thoughts? I maybe be able to try on an esp
> 	equipped ultra1 to see if it has similar issues.
> 

I saw plenty of reports that combination of esp and tagged queueing
caused exactly this failure. I don't know what it is, driver lossage,
defective hardrive, as I've never experienced it myself. As an experiment
I'd suggest to try to disable tagged queueing (man esp(4)).
	Andrey

matthew green | 4 Apr 04:05 2004
Picon

re: install on 8MB boxes


   > I've been able to install current from netboot on a 8MB sparc IPX, the only
   > requirement is to use a smaller kernel than GENERIC or GENERIC_SCSI3.
   > GENERIC doesn't load, it fails with:
   > Booting netbsd
   > Cannot load netbsd: error=27
   > (27 is EFBIG)
   >
   > I don't think the problem is the amount of RAM, but it's probably the
   > fact that it's 2 segments of 4MB with a hole between them.
   > The kernel I used is just GENERIC with sun4m-specific things commented out,
   > no RASTERCONSOLE, no KTRACE/SYSTRACE, no pseudo-filesystems (they can be
   > loaded as LKM), and no network protocols others than INET and INET6.
   > There may be a few more things that could be removed, like sbus cards known
   > to work only on sun4m systems. This kernel should work for all sun4 and sun4c
   > systems, and is 2762499 bytes (vs 3472644).
   >
   > Is it worth adding to the list of standard kernels and distribution,
   > or do we assume peoples with low memory machines will be able to cross-compile
   > their own kernel ?

   	I do not think it is fair to force a new user to cross-compile
   	just to test out NetBSD.

   	Are there any sun4m boxes with less than 8M contiguous memory?
   	The lowest spec sun4m would probably be the classic which requires
   	at a minimum a pair of 4MB simms.

are these 2 simms mapped contiguously?  i dunno :)

(Continue reading)

der Mouse | 4 Apr 07:16 2004
Picon

Re: install on 8MB boxes

>> Are there any sun4m boxes with less than 8M contiguous memory?  The
>> lowest spec sun4m would probably be the classic which requires at a
>> minimum a pair of 4MB simms.
> are these 2 simms mapped contiguously?  i dunno :)

Well, I don't _know_, but machines that require memory to be installed
in sets like pairs usually have a memory bus that's wider than the
memory sticks (like the Sun-3/60, which takes 30-pin SIMMs (which are 8
bits wide) in sets of four - it has a 32-bit memory path).

Thus, I expect the machine has a 64-bit memory path and the SIMMs are
not mapped _contiguously_, exactly; I would say, rather, _interleaved_.

Which, for the purposes at hand, counts as a "yes", since the memory
they provide _is_ contiguous.

/~\ The ASCII				der Mouse
\ / Ribbon Campaign
 X  Against HTML	       mouse <at> rodents.montreal.qc.ca
/ \ Email!	     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B

adrianp | 4 Apr 10:06 2004
Picon

-current build problems

I pulled down -current from CVS yesterday (3/4) and it failed with the
attached error.

Is it related to this ?

http://mail-index.netbsd.org/source-changes/2004/04/03/0029.html

thanks,

adrian.

#   compile  MRCOFFEE_OFW/fpu_subr.o
/usr/src/obj/tooldir.NetBSD-1.6.2-i386/bin/sparc--netbsdelf-gcc -mno-fpu
-ffreestanding -pipe -mcpu=supersparc -O2 -Werror -Wall -Wno-main
-Wno-format-zero-length -Wpointer-arith -Wmissing-prototypes
-Wstrict-prototypes -Wno-sign-compare -I. -I/usr/src/sys/arch
-I/usr/src/sys -nostdinc -DLKM -DPROM_AT_F0 -DMAXUSERS=32 -D_KERNEL
-D_KERNEL_OPT -c /usr/src/sys/arch/sparc/fpu/fpu_subr.c
#   compile  MRCOFFEE_OFW/autoconf.o
/usr/src/obj/tooldir.NetBSD-1.6.2-i386/bin/sparc--netbsdelf-gcc -mno-fpu
-ffreestanding -pipe -mcpu=supersparc -O2 -Werror -Wall -Wno-main
-Wno-format-zero-length -Wpointer-arith -Wmissing-prototypes
-Wstrict-prototypes -Wno-sign-compare -I. -I/usr/src/sys/arch
-I/usr/src/sys -nostdinc -DLKM -DPROM_AT_F0 -DMAXUSERS=32 -D_KERNEL
-D_KERNEL_OPT -c /usr/src/sys/arch/sparc/sparc/autoconf.c
/usr/src/sys/arch/sparc/sparc/autoconf.c: In function `find_cpus':
/usr/src/sys/arch/sparc/sparc/autoconf.c:177: warning: unused variable
`unode'

*** Failed target:  autoconf.o
(Continue reading)

peter waterland | 4 Apr 20:16 2004
Picon

Any issues with qec on a sparc 10?

I was wondering if anyone has had any problems with getting QUAD ETHERNET 
CONTROLLER sbus card (qec) working with a sparc 10? Although it doesnt seem 
to be supported - would it work in an ipx or ipc?

Thanks in advance,

Pete.

_________________________________________________________________
Stay in touch with absent friends - get MSN Messenger 
http://www.msn.co.uk/messenger

der Mouse | 4 Apr 20:53 2004
Picon

Re: Any issues with qec on a sparc 10?

> I was wondering if anyone has had any problems with getting QUAD
> ETHERNET CONTROLLER sbus card (qec) working with a sparc 10?
> Although it doesnt seem to be supported - would it work in an ipx or
> ipc?

I've got a qec.  I've been using it in an IPX for ages and ages (years,
at least), so I can certainly certify that at least one <qec,IPX> pair
works.

I've never tried it in anything else; it's a fairly central house
machine, so I'd rather not start swapping things around to test.

/~\ The ASCII				der Mouse
\ / Ribbon Campaign
 X  Against HTML	       mouse <at> rodents.montreal.qc.ca
/ \ Email!	     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B


Gmane