Messages in this thread | | | Date | Tue, 29 Jul 1997 00:56:52 -0400 (EDT) | From | Jon Lewis <> | Subject | ppp bug? |
| |
On Mon, 28 Jul 1997, Theodore Y. Ts'o wrote:
> As far as not responding to e-mail, please understand that I get around > 800-1200 pieces of e-mail a day, or approximately (1.5 megabytes/day). > I try to respond to as many as I can, but I don't work on Linux
I thought I was bad with about half that number of messages. I don't know how you can read that many a day.
I just noticed the following gp on my 80 port rocketport box. Fortunately, it seems to have been "mostly harmless" as it didn't bring the system down and happened a few days ago.
general protection: 0000 CPU: 0 EIP: 0010:[<02834805>] EFLAGS: 00010202 eax: f000e3d4 ebx: 00000000 ecx: 00000000 edx: 00cc1024 esi: 00cc1024 edi: 0000000e ebp: 00000000 esp: 00c9fe98 ds: 0018 es: 0018 fs: 002b gs: 002b ss: 0018 Process pppd (pid: 18986, process nr: 91, stackpage=00c9f000) Stack: 00000000 00000000 00000020 0000000e 00000216 00000000 00cc1024 ffff0af7 02834941 00cc1024 00000000 00000000 ffff000a 02835b4c 00cc1024 00000000 00cc1024 00cc1024 00de1578 0000c021 0000c021 00000003 02835c91 00cc1024 Call Trace: [<02834941>] [<02835b4c>] [<02835c91>] [<02835f58>] [<00175a95>] [<02835d30>] [<0012110b>] [<0010a730>] [<0010a5a5>] Code: ff d0 83 c4 10 85 c0 7d 0a ff 86 a0 00 00 00 89 f8 eb 03 01
ewok:~# ksymoops /kernels/System.map-2.0.27-fdt <oops.7-25-97 Using `/kernels/System.map-2.0.27-fdt' to map addresses to symbols.
Trace: 2834941 Trace: 2835b4c Trace: 2835c91 Trace: 2835f58 Trace: 175a95 <tty_write+dd/130> Trace: 2835d30 Trace: 12110b <sys_write+10f/148> Trace: 10a730 <error_code+40/50> Trace: 10a5a5 <system_call+55/80>
Code: call *%eax Code: addl $0x10,%esp Code: testl %eax,%eax Code: jnl 00000013 <_EIP+13> Code: incl 0xa0(%esi) Code: movl %edi,%eax Code: jmp 00000016 <_EIP+16> Code: addl %eax,(%eax) Code: nop Code: nop Code: nop
The following modules are in use: Module: #pages: Used by: bsd_comp 1 32 ppp 5 [bsd_comp] 43 slip 2 0 slhc 2 [ppp slip] 43 rocket 5 80
Using a system.map that I reconstructed using the original System.map and the map files I got when the modules were loaded with insmod -m, I get this:
>>EIP: 2834805 <ppp_tty_wakeup_code+41/f4> Trace: 2834941 <ppp_kick_tty+55/5c> Trace: 2835b4c <ppp_dev_xmit_lower+4e4/4f0> Trace: 2835c91 <ppp_dev_xmit_frame+139/144> Trace: 2835f58 <ppp_tty_write+228/240> Trace: 175a95 <tty_write+dd/130> Trace: 2835f58 <ppp_tty_write+228/240> Trace: 12110b <sys_write+10f/148> Trace: 10a730 <error_code+40/50> Trace: 10a5a5 <system_call+55/80>
Code: 2834805 <ppp_tty_wakeup_code+41/f4> call *%eax Code: 2834807 <ppp_tty_wakeup_code+43/f4> addl $0x10,%esp Code: 283480a <ppp_tty_wakeup_code+46/f4> testl %eax,%eax Code: 283480c <ppp_tty_wakeup_code+48/f4> jnl 2834818 <ppp_tty_wakeup_code+54/f4> Code: 283480e <ppp_tty_wakeup_code+4a/f4> incl 0xa0(%esi) Code: 2834814 <ppp_tty_wakeup_code+50/f4> movl %edi,%eax Code: 2834816 <ppp_tty_wakeup_code+52/f4> jmp 283481b <ppp_tty_wakeup_code+57/f4> Code: 2834818 <ppp_tty_wakeup_code+54/f4> addl %eax,(%eax) Code: 283481a <ppp_tty_wakeup_code+56/f4> nop Code: 283481b <ppp_tty_wakeup_code+57/f4> nop Code: 283481c <ppp_tty_wakeup_code+58/f4> nop
Any idea what actually happened? The GP happened before any of the normal syslog()s for pppd starting up. In fact there were none of the expected pppd messages for pid 18986...so I assume it happened very early in the startup phase of a PPP session. The system has been exceptionally stable, so suggestions of bad RAM >/dev/null :). It's a Tomcat I, 32mb true parity memory, P100, 2 RA32, 1 RA16, Trident 9440 PCI VGA, NCR810 SCSI, 3c509b. pppd is 2.2.0f. I'm using rocket.o 1.12 with a few patches from Doug Ledford. Should I use 1.14 when I do the upgrade? Is there a reason it's still considered ALPHA 6 months after it was released?
Linux 2.0.27 (root@sloth) (gcc 2.7.2.1) #4 Thu Dec 26 13:15:46 EST 1996 [ewok]
Memory: Total Used Free Shared Buffers Cached Mem: 31012 29296 1716 4676 224 1716 Swap: 60660 10468 50192
Bootup: Mon Jan 20 11:49:47 1997 Load average: 1.03 1.21 1.37 2/105 2185
user : 20:18:29.50 0.4% page in : 146592921 disk 1: 47473470r55746316w nice : 0:00:00.00 0.0% page out: 65874182 system: 2d 4:51:51.66 1.2% swap in : 7397474 idle : 186d 10:17:39.09 98.4% swap out: 1809851 uptime: 189d 11:28:00.24 context : 539683235
irq 0: 1637088025 timer irq 8: 0 + rtc irq 1: 490 keyboard irq 9: 0 irq 2: 0 cascade irq 10: 654067998 3c509 irq 3: 0 irq 11: 104625480 + 53c8xx irq 4: 0 irq 12: 0 irq 5: 0 irq 13: 1 math error irq 6: 2 irq 14: 0 irq 7: 0 irq 15: 0
With the 80 ports, as well as gated, caching named, and sshd, the system really doesn't have enough RAM anymore, and I'm hoping to upgrade the kernel and throw another 16mb into it sometime soon. The question is 2.0.29, 2.0.30, or 2.0.31-pre-n? It'll be a shame to let all that uptime escape :)
I have a system (endor) identical to ewok in every way (same hardware and kernel/modules binaries) except that it has 64 Cyclades ports instead of 80 Rocketports, and its uptime is 9 days longer. Nothing like this has happened to endor yet. ------------------------------------------------------------------ Jon Lewis <jlewis@fdt.net> | Unsolicited commercial e-mail will Network Administrator | be proof-read for $199/message. Florida Digital Turnpike | ________Finger jlewis@inorganic5.fdt.net for PGP public key_______
| |