[parisc-linux] strace locks up 2.4.22-pa14

[parisc-linux] strace locks up 2.4.22-pa14

[Grant Grundler]

In trying to debug a problem with dcopserver not starting,
I tried: 
	strace -f -o /tmp/dcop.out dcopserver

This locked up the system. I could still ping the tg3 LAN (interrupts
are enabled?). But Mouse and keyboard stopped responding.
I TOC'd the system - but forgot to dump PIM on the way up.
Since it's a graphical console, I don't plan on copying down
all the info.  If anyone wants to see that, please let me know
which registers are of interest (IOAQ, gr02, gr26-23 are obvious).

grant

RE: [parisc-linux] strace locks up 2.4.22-pa14

[Joel Soete]

Hi Grant,

I am tring to track a lpr pb on my b2k and so try:
# strace -f /usr/bin/lpr foo.doc

that works;

I read your post and try so now:
# strace -f -o /var/tmp/Strace.doc /usr/bin/lpr foo.doc

and system also hang (same behaviour as) and so I toc it (which seems to
failled ; apparently because of ide-cdrom???)

I took the opportunity to unplug ide-cable's cdrom (to test various).

Then to try to obtain a relevant toc, I re-launch:
# strace -f -o /var/tmp/Strace.doc /usr/bin/lpr foo.doc

This time it works??
Is it the same for you after the reboot?
(sorry, I tried severall time strace but no more means to reproduce)

J.

PS: btw the end of my strace is:
[...]
24801 write(3, "\1lp\n", 4)             = 4
24801 read(3, "\0", 1)                  = 1
24801 close(3)                          = 0
24801 SYS_222(0, 0, 0x1, 0x2f, 0)       = -1 ENOSYS (Function not implemented)
24801 exit(0)                           = ?

And at the same time, at the serial console I got:
+ set -v
+ exec

Have you any advise about what I miss?
>-- Original Message --
>From: Grant Grundler <grundler@parisc-linux.org>
>To: parisc-linux@lists.parisc-linux.org
>Date: Thu, 20 Nov 2003 17:50:03 -0700
>Subject: [parisc-linux] strace locks up 2.4.22-pa14
>
>
>
In trying to debug a problem with dcopserver not starting,
I tried: 
	strace -f -o /tmp/dcop.out dcopserver

This locked up the system. I could still ping the tg3 LAN (interrupts
are enabled?). But Mouse and keyboard stopped responding.
I TOC'd
>the system - but forgot to dump PIM on the way up.
Since it's a graphical console, I don't plan on copying down
all the info.  If anyone wants to see that, please let me know
which registers are of interest (IOAQ, gr02, gr26-23 are obvious).

gran
>
_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux



-------------------------------------------------------------------------
Tiscali ADSL: 12 mois à 29,50 €/mois! L'Internet rapide, c'est pour tout
le monde.
http://reg.tiscali.be/default.asp?lg=fr

Re: [parisc-linux] strace locks up 2.4.22-pa14

[Grant Grundler]

On Fri, Nov 21, 2003 at 01:39:49PM +0100, Joel Soete wrote:
> Hi Grant,
> 
> I am tring to track a lpr pb on my b2k and so try:
> # strace -f /usr/bin/lpr foo.doc
> 
> that works;

ok

> I read your post and try so now:
> # strace -f -o /var/tmp/Strace.doc /usr/bin/lpr foo.doc
> 
> and system also hang (same behaviour as)
> and so I toc it (which seems to failled ; apparently because of ide-cdrom???)

hrm. I doubt the ide-cdrom is causing the TOC to fail. 
It's likely the CD-ROM drive is getting reset when TOC occurs
and that's the noise you hear.
Something else might be causing problems.

> I took the opportunity to unplug ide-cable's cdrom (to test various).
> 
> Then to try to obtain a relevant toc, I re-launch:
> # strace -f -o /var/tmp/Strace.doc /usr/bin/lpr foo.doc
> 
> This time it works??
> Is it the same for you after the reboot?

I haven't retried because I need that to actaully work on that system.
I'm building a new kernel and will try at home tonight.


> (sorry, I tried severall time strace but no more means to reproduce)

bummer. Makes me wonder if this is CPU cache related too. 

> PS: btw the end of my strace is:
> [...]
> 24801 write(3, "\1lp\n", 4)             = 4
> 24801 read(3, "\0", 1)                  = 1
> 24801 close(3)                          = 0
> 24801 SYS_222(0, 0, 0x1, 0x2f, 0)       = -1 ENOSYS (Function not implemented)
> 24801 exit(0)                           = ?
> 
> And at the same time, at the serial console I got:
> + set -v
> + exec
> 
> Have you any advise about what I miss?

include/asm/unistd.h says:
#define __NR_exit_group         (__NR_Linux + 222)

No architecture implements that syscall.
("find -name '*.[S]' | xargs fgrep exit_group")
I have no clue what it's supposed to do and looks like
I don't need to care either.

Perhaps just recompiling the application or linked lib that uses
that syscall might get rid of it.

grant

Re: [parisc-linux] strace locks up 2.4.22-pa14

[Joel Soete]

Grant Grundler wrote:
> On Fri, Nov 21, 2003 at 01:39:49PM +0100, Joel Soete wrote:
> 
>>Hi Grant,
>>
>>I am tring to track a lpr pb on my b2k and so try:
>># strace -f /usr/bin/lpr foo.doc
>>
>>that works;
> 
> 
> ok
> 
> 
>>I read your post and try so now:
>># strace -f -o /var/tmp/Strace.doc /usr/bin/lpr foo.doc
>>
>>and system also hang (same behaviour as)
>>and so I toc it (which seems to failled ; apparently because of ide-cdrom???)
> 
> 
> hrm. I doubt the ide-cdrom is causing the TOC to fail. 

I don't know but seems to hang like boot hang when I include the pdc 
support into the kernel: green led stay lightning. (and toc seems to 
failed because of a time out IO; i don't remember but I think i do have 
to read it on lcd)

> It's likely the CD-ROM drive is getting reset when TOC occurs
> and that's the noise you hear.
> Something else might be causing problems.
> 
Well, I don't realy need that system hang to toc it? (just think right now)
So when I have some time, I will re-try a toc when system is running and 
see what hapen ;)
> 
>>I took the opportunity to unplug ide-cable's cdrom (to test various).
>>
>>Then to try to obtain a relevant toc, I re-launch:
>># strace -f -o /var/tmp/Strace.doc /usr/bin/lpr foo.doc
>>
>>This time it works??
>>Is it the same for you after the reboot?
> 
> 
> I haven't retried because I need that to actaully work on that system.
> I'm building a new kernel and will try at home tonight.
> 
Ok
> 
> 
>>(sorry, I tried severall time strace but no more means to reproduce)
> 
> 
> bummer. Makes me wonder if this is CPU cache related too. 
> 
I am curious to see the result of your test ?)
> 
>>PS: btw the end of my strace is:
>>[...]
>>24801 write(3, "\1lp\n", 4)             = 4
>>24801 read(3, "\0", 1)                  = 1
>>24801 close(3)                          = 0
>>24801 SYS_222(0, 0, 0x1, 0x2f, 0)       = -1 ENOSYS (Function not implemented)
>>24801 exit(0)                           = ?
>>
>>And at the same time, at the serial console I got:
>>+ set -v
>>+ exec
>>
>>Have you any advise about what I miss?
> 
> 
> include/asm/unistd.h says:
> #define __NR_exit_group         (__NR_Linux + 222)
> 
> No architecture implements that syscall.
> ("find -name '*.[S]' | xargs fgrep exit_group")
> I have no clue what it's supposed to do and looks like
> I don't need to care either.
> 
> Perhaps just recompiling the application or linked lib that uses
> that syscall might get rid of it.
> 
Good idea, I will try.

what I don't understand is that lpr works fine on my b180 runing a 
k-2.4.21 (for lvm2 support) otc the b2k is running a k-2.4.23-rc2-pa7 (a 
merge of mine); so my doubt is that I could break something :(.

Thanks for advise,
	Joel


> grant
> 

Re: [parisc-linux] strace locks up 2.4.22-pa14

[Grant Grundler]

On Sat, Nov 22, 2003 at 11:58:29AM +0000, Joel Soete wrote:
> I don't know but seems to hang like boot hang when I include the pdc 
> support into the kernel: green led stay lightning. (and toc seems to 
> failed because of a time out IO; i don't remember but I think i do have 
> to read it on lcd)

What do you mean by "PDC support"?
You mean CONFIG_PDC_CHASSIS?
AFAIK, this only enables chassis codes to be displayed on either
the "front panel" LCD or LEDs.

It looks like my box HPMCs after the TOC. The TOC performs
a reset and then seltests run. Then next thing on the LCD
display is "OS HPMC" even though the gfx card hasn't been
initialized yet.

> Well, I don't realy need that system hang to toc it?

Right. TOC is just convenient to use when the system is hung.

> So when I have some time, I will re-try a toc when system is running and 
> see what hapen ;)

You'll get state info about cpu_idle() or whatever it happens to
be doing.

> >bummer. Makes me wonder if this is CPU cache related too. 
> I am curious to see the result of your test ?)

2.4.22-pa17 on c3k, "strace -f -o /tmp/dcop.out dcopserver" hung at home. 

GR02 0x10389567	
20-23	0x32  0x0e 0x10 0x00
24-27   0x00  0x10 0xcc 0x102d2010

CR0-3 0x00 0x1822 0x00 0x1824

IOAQ 0x102a1188	number+170
CPU State 0x9e000001

Looks like the PIM dump is garbage. (gr02 is not a kernel address)
I've cleared the PIM and will retry later...gotta run.


> what I don't understand is that lpr works fine on my b180 runing a 
> k-2.4.21 (for lvm2 support) otc the b2k is running a k-2.4.23-rc2-pa7 (a 
> merge of mine); so my doubt is that I could break something :(.

Is this via parallel port or network printer?
If network, then I would guess your merge (or the 2.4.23-rc2 base) has a bug.
If parallel port, likely a bug in suckyio support for parallel port.

But it's not exactly "apples to apples" comparison since the IO subsystem
between B180 and B2k is completely different.

grant

Re: [parisc-linux] strace locks up 2.4.22-pa14

[Joel Soete]

Grant Grundler wrote:
> On Sat, Nov 22, 2003 at 11:58:29AM +0000, Joel Soete wrote:
> 
>>I don't know but seems to hang like boot hang when I include the pdc 
>>support into the kernel: green led stay lightning. (and toc seems to 
>>failed because of a time out IO; i don't remember but I think i do have 
>>to read it on lcd)
> 
> 
> What do you mean by "PDC support"?
> You mean CONFIG_PDC_CHASSIS?

Sorry I will have to check ((not the box near hand :( )

> AFAIK, this only enables chassis codes to be displayed on either
> the "front panel" LCD or LEDs.
> 
> It looks like my box HPMCs after the TOC.
Yes

> The TOC performs
> a reset
Yes (i trust ;) )
> and then seltests run.
Never for me, stay lock with ide-cdrom led lightning

> Then next thing on the LCD
> display is "OS HPMC"

Yes and on the half botum part are wraping some additional text

> even though the gfx card hasn't been
> initialized yet.
> 
I don't select sti support (builtin gfx not supported on b2k)
> 
>>Well, I don't realy need that system hang to toc it?
> 
> 
> Right. TOC is just convenient to use when the system is hung.
> 
> 
>>So when I have some time, I will re-try a toc when system is running and 
>>see what hapen ;)
> 
> 
> You'll get state info about cpu_idle() or whatever it happens to
> be doing.
> 
> 
>>>bummer. Makes me wonder if this is CPU cache related too. 
>>
>>I am curious to see the result of your test ?)
> 
> 
> 2.4.22-pa17 on c3k, "strace -f -o /tmp/dcop.out dcopserver" hung at home. 
> 
hmm b2k and c3k are very brother system in term of architecture and 
components? May I ask you the amount of ram stand in your c3k? (on my 
b2k only 256k).

> GR02 0x10389567	
> 20-23	0x32  0x0e 0x10 0x00
> 24-27   0x00  0x10 0xcc 0x102d2010
> 
> CR0-3 0x00 0x1822 0x00 0x1824
> 
> IOAQ 0x102a1188	number+170
> CPU State 0x9e000001
> 
> Looks like the PIM dump is garbage. (gr02 is not a kernel address)
> I've cleared the PIM and will retry later...gotta run.
> 
hmm the same when I tried to analyse hpmc gr02=0x00000150?

> 
> 
>>what I don't understand is that lpr works fine on my b180 runing a 
>>k-2.4.21 (for lvm2 support) otc the b2k is running a k-2.4.23-rc2-pa7 (a 
>>merge of mine); so my doubt is that I could break something :(.
> 
> 
> Is this via parallel port or network printer?
Sorry, via network (hp lazerjet 4mv)

> If network, then I would guess your merge (or the 2.4.23-rc2 base) has a bug.

Certainly :( [but where, the rest ftp, rsync, nfs-client, gtk-app, ... 
work fine and no special messages neither into kernel, messages or 
syslog?]. Well don't care now, it is just a beat anoying to printout my 
scripts.

> If parallel port, likely a bug in suckyio support for parallel port.
> 
> But it's not exactly "apples to apples" comparison since the IO subsystem
> between B180 and B2k is completely different.
> 
agree

Thanks,
	Joel

> grant
> 

Re: [parisc-linux] strace locks up 2.4.22-pa14

[Grant Grundler]

On Sat, Nov 22, 2003 at 10:45:51PM +0000, Joel Soete wrote:
> hmm b2k and c3k are very brother system in term of architecture and 
> components?

yes - identical nearly. Share the same CPU, chipset, and firmware.

>  May I ask you the amount of ram stand in your c3k? (on my 
> b2k only 256k).

1GB.


> >If network, then I would guess your merge (or the 2.4.23-rc2 base) has a 
> >bug.
> 
> Certainly :( [but where, the rest ftp, rsync, nfs-client, gtk-app, ... 
> work fine and no special messages neither into kernel, messages or 
> syslog?]. Well don't care now, it is just a beat anoying to printout my 
> scripts.

ok. TBH, I have no clue then. It might be possible to attach strace to
a running lp deamon, redirect strace output to /dev/console and then
compore what happens different on the b2k when a print job is started.
My hope is the syscall output will get printed before it totallte crashes.

grant

Re: [parisc-linux] strace locks up 2.4.22-pa14

[Joel Soete]

>>  May I ask you the amount of ram stand in your c3k? (on my 
>> b2k only 256k).
>
>1GB.
>
I would be curious if the same behavior occurs with only 512K?

>> >If network, then I would guess your merge (or the 2.4.23-rc2 base) has
a 
>> >bug.
>> 
>> Certainly :( [but where, the rest ftp, rsync, nfs-client, gtk-app, ...

>> work fine and no special messages neither into kernel, messages or 
>> syslog?]. Well don't care now, it is just a beat anoying to printout my

>> scripts.
>
>ok. TBH, I have no clue then. It might be possible to attach strace to
>a running lp deamon, redirect strace output to /dev/console and then
>compore what happens different on the b2k when a print job is started.

Good idea, I will try to see howto.

>My hope is the syscall output will get printed before it totallte crashes.

That should be a good explanation.

Thanks for attention,
Joel



-------------------------------------------------------------------------
Tiscali ADSL: 12 mois à 29,50 €/mois! L'Internet rapide, c'est pour tout
le monde.
http://reg.tiscali.be/default.asp?lg=fr

Re: [parisc-linux] strace locks up 2.4.22-pa14

[Grant Grundler]

On Mon, Nov 24, 2003 at 08:21:28AM +0100, Joel Soete wrote:
> I would be curious if the same behavior occurs with only 512K?

sorry - I don't think it's worth removing 512 MB to try.

grant

Re: [parisc-linux] strace locks up 2.4.22-pa14

[Joel Soete]

Hi Grant,

Just a follow up:

>-- Original Message --
>From: "Joel Soete" <soete.joel@tiscali.be>
>To: "Grant Grundler" <grundler@parisc-linux.org>
>Cc: parisc-linux@lists.parisc-linux.org
>Date: Mon, 24 Nov 2003 08:21:28 +0100
>Subject: Re: [parisc-linux] strace locks up 2.4.22-pa14
>
>
>

>>ok. TBH, I have no clue then. It might be possible to attach strace to
>>a running lp deamon, redirect strace output to /dev/console and then
>>compore what happens different on the b2k when a print job is started.
>
>Good idea, I will try to see howto
>
Oops, I try first to strace lpd ("strace -f -p xxx 2>&1 | tee Strace-lpd.log")on
the b180 (running a 2.4.19 + lvm (aka dm) support) but this "locked up" too.

Unfortunately, I don't have anymore a 'woody' install to try to trace back
when the strace pb appears :(.

btw since my last update (last friday), I couldn't any more print with this
b180 too.

J.


-------------------------------------------------------------------------
Tiscali ADSL: 12 mois à 29,50 €/mois! L'Internet rapide, c'est pour tout
le monde.
http://reg.tiscali.be/default.asp?lg=fr

Re: [parisc-linux] strace locks up 2.4.22-pa14

[Joel Soete]

Grant Grundler wrote:
> On Sat, Nov 22, 2003 at 11:58:29AM +0000, Joel Soete wrote:
> 
>>I don't know but seems to hang like boot hang when I include the pdc 
>>support into the kernel: green led stay lightning. (and toc seems to 
>>failed because of a time out IO; i don't remember but I think i do have 
>>to read it on lcd)
> 
> 
> What do you mean by "PDC support"?
> You mean CONFIG_PDC_CHASSIS?

Sorry that was in an old mail 
<http://lists.parisc-linux.org/pipermail/parisc-linux/2002-December/018689.html>
and much concern CONFIG_PDC_CONSOLE (via CONFIG_SERIAL_NONSTANDARD=y)


thanks for your attention,
	Joel

Re: [parisc-linux] strace locks up 2.4.22-pa14

[Grant Grundler]

On Sat, Nov 22, 2003 at 10:49:20PM +0000, Joel Soete wrote:
> >2.4.22-pa17 on c3k, "strace -f -o /tmp/dcop.out dcopserver" hung at home. 
> Sorry I forget to ask if "strace -f dcopserver" works?

good question. It locked up with 2.4.22-pa17 as well. :^(
"locked up" meaning:
o I could ping a NIC
o console mouse/keyboard stopped responding
o I couldn't ssh into the box.

System TOC'd nicely and then HPMC'd the box.
GR00-3 0x0  0x1032f010 0x102a1a70  0x1038953c
GR20-23 0x53  0xff..ff 0x0a  0x3c230a1c
GR24-27  0x102bddbd 0xff...feb 0x3c230948 0x102d2010

IIA Space  0x0
IIA Offset 0x102a0994 strnlen+18
GR02 0x102a1a70 vsnprintf+6dc

blecchh...need a full stack unwinder.

grant

Re: [parisc-linux] strace locks up 2.4.22-pa14

[Joel Soete]

>"locked up" meaning:
>o I could ping a NIC
>o console mouse/keyboard stopped responding
>o I couldn't ssh into the box.
>
Exactly what I see when i reach to reproduce

>System TOC'd nicely and then HPMC'd the box.
>GR00-3 0x0  0x1032f010 0x102a1a70  0x1038953c
>GR20-23 0x53  0xff..ff 0x0a  0x3c230a1c
>GR24-27  0x102bddbd 0xff...feb 0x3c230948 0x102d2010
>
>IIA Space  0x0
>IIA Offset 0x102a0994 strnlen+18
>GR02 0x102a1a70 vsnprintf+6dc
Exactly the same for the 'strace lpr' (excepted address)?
grrr I retry this morning, no more means to reproduce :(

But you have right for the toc (removing cdrom doesn't help).
I just trigger and notice 4 messages wraping on lcd:
FLT CBF0: SYS DB
HPMC Initiated

FLT 500B: SYS DB
Bus time out

FLT CBF1: SYS DB
no OS HPMC IVA

FLT CBFC: SYS DB
OS HPMC br err

And only power off / on allow to restart. (id don't remember if toc sync
disk or not, but I noticed also that fs have to fsck)

OTC it help to boot 2.4.22-pa12 + aty patch :) (i will see what can I do
with that stuff)

>blecchh...need a full stack unwinder.

Courage ;),
    Joel


-------------------------------------------------------------------------
Tiscali ADSL: 12 mois à 29,50 €/mois! L'Internet rapide, c'est pour tout
le monde.
http://reg.tiscali.be/default.asp?lg=fr

Re: [parisc-linux] strace locks up 2.4.22-pa14

[Grant Grundler]

On Mon, Nov 24, 2003 at 08:49:20AM +0100, Joel Soete wrote:
> And only power off / on allow to restart. (id don't remember if toc sync
> disk or not, but I noticed also that fs have to fsck)

TOC has nothing to do with the file system

grant