2.1.109: TR Broken?

2.1.109: TR Broken?

Post by bill davids » Fri, 24 Jul 1998 04:00:00



I just tried 2.1.109 for the first time, on a machine which has been
happily running 2.1.101 since that was new. The machine hangs in the
token ring driver install, and alt-cntl-del is the only way out.

Since I depend on the TR to supply much of what the machine does, I may
not be able to debug this very effectively. If someone else gets the
urge to look at this, please do. I'm not going to NFS mount my debugging
tools, and I see that the driver has been hacked enhanced to output the
MTU for both 16Mbit and 4Mbit speeds. Undoubtedly that was why the
change was put in during a bug freeze, um... code freeze.

Stock IBM TR 4/16 auto, init with isapnp to standard defaults, irq 10, c800.
--

"If heros often had feet of clay, they were, at least, still heros. How
many other clay-footed humans could claim as much?"
        Michael Flynn, _Firestar_

 
 
 

2.1.109: TR Broken?

Post by Paul Norto » Fri, 24 Jul 1998 04:00:00



> I just tried 2.1.109 for the first time, on a machine which has been
> happily running 2.1.101 since that was new. The machine hangs in the
> token ring driver install, and alt-cntl-del is the only way out.

Is the driver generated as part of the kernel or as a module?

Quote:> Since I depend on the TR to supply much of what the machine does, I may
> not be able to debug this very effectively. If someone else gets the
> urge to look at this, please do. I'm not going to NFS mount my debugging
> tools, and I see that the driver has been hacked enhanced to output the
> MTU for both 16Mbit and 4Mbit speeds. Undoubtedly that was why the
> change was put in during a bug freeze, um... code freeze.

MTU, as well as fixing tcpdump and getting rid of 802.2TR and EtherII
for 802.2 and SNAP IPX respectively.

Did you get any tr0 messages in syslog, or did it die without complaint?

> Stock IBM TR 4/16 auto, init with isapnp to standard defaults, irq 10, c800.
> --

> "If heros often had feet of clay, they were, at least, still heros. How
> many other clay-footed humans could claim as much?"
>    Michael Flynn, _Firestar_


 
 
 

2.1.109: TR Broken?

Post by bill davids » Sat, 25 Jul 1998 04:00:00





|
| > I just tried 2.1.109 for the first time, on a machine which has been
| > happily running 2.1.101 since that was new. The machine hangs in the
| > token ring driver install, and alt-cntl-del is the only way out.
|
| Is the driver generated as part of the kernel or as a module?

As a module. I initialize the PnP config with isapnp and then load the
module.

| Did you get any tr0 messages in syslog, or did it die without complaint?

The system doesn't boot, so that's hard to say. It will print an error
message all day, but nothing will move it past that point. Let's see...

Before booting the broken kernel I checked the syslog, nothing there.
But when I boot 2.1.109 I see:
  tr0: open failed: ret_code = 30, retrying

More later... I going to play with it a bit.

--

"If heros often had feet of clay, they were, at least, still heros. How
many other clay-footed humans could claim as much?"
        Michael Flynn, _Firestar_

 
 
 

2.1.109: TR Broken?

Post by Paul Norto » Sat, 25 Jul 1998 04:00:00



>   tr0: open failed: ret_code = 30, retrying

I'll bet we're trying to allocate too much in the
way of send/receive buffers.  How much shared RAM
is the adapter set for?
 
 
 

2.1.109: TR Broken?

Post by bill davids » Tue, 28 Jul 1998 04:00:00





| >   tr0: open failed: ret_code = 30, retrying
|
| I'll bet we're trying to allocate too much in the
| way of send/receive buffers.  How much shared RAM
| is the adapter set for?

After looking at my Friday night typing, I'm surprised you could read
the last post...

The size is 16k, and as noted 2.1.101 works on this machine. Here's the
init file, trimmed of some of the comments. Card 1 is a sound card, and
is totally commented out, since I don't need/want it at the moment.

# Card 2: (serial identifier f5 00 80 99 51 01 00 4d 24)
# IBM0001 Serial No 8427857 [checksum f5]
# Version 1.0, Vendor version 0.1
# Vendor defined tag:  73 01 00 12
# ANSI string -->IBM Auto 16/4 Token-Ring ISA Adapter
#
# Logical device id IBM0000
#     Device support I/O range check register

(CONFIGURE IBM0001/8427857 (LD 0
 (IO 0 (BASE 0x0a20))
 (MEM 0 (BASE 0x0c8000) (MODE wr) (UPPER 0x004000))
 (MEM 1 (BASE 0x0cc000) (MODE wr) (UPPER 0x002000))
 (INT 0 (IRQ 10 (MODE +E)))
 (ACT Y)
))
# End tag... Checksum 0x00 (OK)

With 2.1.101 I see:
tr0: Adaptor initialized and opened.

Oh, 2.1.110 and 2.1.111 don't work, either. I ran about 12 hours of
2.1.111 on a non-TR SMP machine over the weekend, it's normally a demo
machine but currently sitting in as an emergency backup router :-(

--

  What do you mean I shouldn't do things like that at my age? At my
age if you don't do things like that you might die of natural
causes or boredom!

 
 
 

2.1.109: TR Broken?

Post by Paul Norto » Tue, 28 Jul 1998 04:00:00



> The size is 16k, and as noted 2.1.101 works on this machine. Here's the
> init file, trimmed of some of the comments. Card 1 is a sound card, and
> is totally commented out, since I don't need/want it at the moment.

Could you apply this patch to drivers/net/ibmtr.c and see if it helps
any?

--- ibmtr.c.dist        Mon Jul 27 12:38:36 1998

                ti->dhb_size4mb  = MIN(ti->dhb_size4mb, 4464);
                ti->rbuf_len4 = 512;
                ti->rbuf_cnt4 = 9;
-               ti->dhb_size16mb = MIN(ti->dhb_size16mb, 4096);
-               ti->rbuf_len16 = 2048;
-               ti->rbuf_cnt16 = 2;
+               ti->dhb_size16mb = MIN(ti->dhb_size16mb, 3072);
+               ti->rbuf_len16 = 1024;
+               ti->rbuf_cnt16 = 3;
                break;
        case  64 : /* 32KB shared RAM */
                ti->dhb_size4mb  = MIN(ti->dhb_size4mb, 4464);

 
 
 

2.1.109: TR Broken?

Post by bill davids » Wed, 29 Jul 1998 04:00:00





>> The size is 16k, and as noted 2.1.101 works on this machine. Here's the
>> init file, trimmed of some of the comments. Card 1 is a sound card, and
>> is totally commented out, since I don't need/want it at the moment.

>Could you apply this patch to drivers/net/ibmtr.c and see if it helps
>any?

I have, but it doesn't solve the problem. Still getting the status 30.

I really appreciate your looking at this, I'm a bit stuck upgrading
until it's fixed.
--

  What do you mean I shouldn't do things like that at my age? At my
age if you don't do things like that you might die of natural
causes or boredom!

 
 
 

2.1.109: TR Broken?

Post by Paul Norto » Thu, 30 Jul 1998 04:00:00


Well, the ret_code 30 definitely points to a problem with receive buffer
allocation.  Could you try changing:

+               ti->dhb_size16mb = MIN(ti->dhb_size16mb, 3072);
+               ti->rbuf_len16 = 1024;
+               ti->rbuf_cnt16 = 3;

to:

+               ti->dhb_size16mb = MIN(ti->dhb_size16mb, 2048);
+               ti->rbuf_len16 = 1024;
+               ti->rbuf_cnt16 = 2;

I'm amazed that it won't allocate 6K (or even 8k) of transmit/receive
buffer space on the adapter with 16K of shared RAM available.  I'm going
to do some more testing on my systems at home, but I don't have an auto
adapter and my results will probably differ somewhat.





> >> The size is 16k, and as noted 2.1.101 works on this machine. Here's the
> >> init file, trimmed of some of the comments. Card 1 is a sound card, and
> >> is totally commented out, since I don't need/want it at the moment.

> >Could you apply this patch to drivers/net/ibmtr.c and see if it helps
> >any?

> I have, but it doesn't solve the problem. Still getting the status 30.

> I really appreciate your looking at this, I'm a bit stuck upgrading
> until it's fixed.
> --

>   What do you mean I shouldn't do things like that at my age? At my
> age if you don't do things like that you might die of natural
> causes or boredom!

 
 
 

1. Badness in softirq.c 109, possible fix

Hello.

Got badness in local_bh_enable because Im still in irq_disabled.
This is caused by local_irq_save() in do_tty_hangup (drivers/char/tty_io.c)
He calls local_irq_save(flags) and then goes to flush_buffer and down
to local_bh_enable.
Do we really need local_irq_save/restore in do_tty_hangup() line 447 ?
Its not safe.
Thanks

Badness in local_bh_enable at kernel/softirq.c:109
Jun 19 13:21:01 spiderman kernel: Call Trace:
Jun 19 13:21:01 spiderman kernel:  [<c0128e63>] local_bh_enable+0x93/0xa0
Jun 19 13:21:01 spiderman kernel:  [<e08ebed1>]
ppp_async_push+0xe1/0x250 [ppp_async]
Jun 19 13:21:01 spiderman kernel:  [<c0173023>] cached_lookup+0x23/0x90
Jun 19 13:21:01 spiderman kernel:  [<e08eb712>]
ppp_asynctty_wakeup+0x32/0x70 [ppp_async]
Jun 19 13:21:01 spiderman kernel:  [<c022297a>] pty_unthrottle+0x5a/0x60
Jun 19 13:21:01 spiderman kernel:  [<c021e2bb>] check_unthrottle+0x3b/0x40
Jun 19 13:21:01 spiderman kernel:  [<c021e370>]
reset_buffer_flags+0xb0/0x100
Jun 19 13:21:01 spiderman kernel:  [<c021e3d3>] n_tty_flush_buffer+0x13/0x60
Jun 19 13:21:01 spiderman kernel:  [<c0222d46>] pty_flush_buffer+0x66/0x70
Jun 19 13:21:01 spiderman pppd[1366]: Modem hangup
Jun 19 13:21:01 spiderman kernel:  [<c021a105>] do_tty_hangup+0x4f5/0x5d0

Jun 19 13:21:01 spiderman pppd[1366]: Connection terminated.
Jun 19 13:21:01 spiderman kernel:  [<c021bbaf>] release_dev+0x6df/0x730
Jun 19 13:21:01 spiderman pppd[1366]: Connect time 9.8 minutes.
Jun 19 13:21:01 spiderman kernel:  [<c014bd8c>] release_pages+0x1dc/0x280
Jun 19 13:21:01 spiderman pppd[1366]: Sent 84 bytes, received 50 bytes.
Jun 19 13:21:01 spiderman kernel:  [<c0150dbb>] zap_pmd_range+0x4b/0x70
Jun 19 13:21:01 spiderman kernel:  [<c017dce2>] dput+0x22/0x380
Jun 19 13:21:01 spiderman kernel:  [<c021bffb>] tty_release+0x3b/0xc0
Jun 19 13:21:01 spiderman kernel:  [<c01640f8>] __fput+0x118/0x120
Jun 19 13:21:01 spiderman kernel:  [<c01624d0>] filp_close+0xd0/0x130
Jun 19 13:21:01 spiderman kernel:  [<c0125f68>] put_files_struct+0x58/0xc0
Jun 19 13:21:01 spiderman kernel:  [<c0126dfe>] do_exit+0x1fe/0x690
Jun 19 13:21:02 spiderman kernel:  [<c01624d0>] filp_close+0xd0/0x130
Jun 19 13:21:02 spiderman kernel:  [<c0127419>] do_group_exit+0xf9/0x180
Jun 19 13:21:02 spiderman kernel:  [<c01625bb>] sys_close+0x8b/0x110
Jun 19 13:21:02 spiderman kernel:  [<c010b1fb>] syscall_call+0x7/0xb

Petur Thors

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in

More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

2. java virtual machine for free bsd!!!

3. Alpha 2.1.109?

4. FTP by script

5. 2.5.66-bk5 spinlock warnings/errors - Specifically ide-io:109 spinlock notice

6. e2fsck -c and badblocks not fixing problem

7. troubles installing linux 109

8. HELP! how do I undelete a file!??

9. linux 109 doesn't install right.

10. No errpt via smit (1800-109, 1800-107)

11. AIX 4.3.3 ML5 : sem_init fails with ENOSYS (errno 109)

12. Cannot compile kernel 2.1.109

13. 2.5.72 Badness in local_bh_enable at kernel/softirq.c:109