kern/124127: [msk] watchdog timeout (missed Tx interrupts) -- recovering

[Pyun YongHyeon]

The following reply was made to PR kern/124127; it has been noted by GNATS.

From: Pyun YongHyeon <pyunyh at gmail.com>
To: Mark Atkinson <atkin901 at yahoo.com>
Cc: freebsd-net at freebsd.org, bug-followup at FreeBSD.org
Subject: Re: kern/124127: [msk] watchdog timeout (missed Tx interrupts) -- recovering
Date: Thu, 29 Oct 2009 09:49:09 -0700

 On Thu, Oct 29, 2009 at 06:52:34AM -0700, Mark Atkinson wrote:
 > Wow, not sure what to blame for that charset nightmare.  Apologies.
 > Here's the original message:
 > 
 > On the unpatched -current kernel, built
 > 
 > FreeBSD hellfire.filament.org 9.0-CURRENT FreeBSD 9.0-CURRENT #14: Mon
 > Oct 19 09:12:03 PDT 2009
 > 
 > I recieved the following panic today related to this:
 > 
 > Fatal trap 12: page fault while in kernel mode
 > cpuid = 0; apic id = 00
 > fault virtual address  = 0xdeadc10a
 > fault code              = supervisor read, page not present
 > instruction pointer    = 0x20:0xc0987410
 > stack pointer          = 0x28:0xd533dac0
 > frame pointer          = 0x28:0xd533dae8
 > code segment            = base 0x0, limit 0xfffff, type 0x1b
 >                         = DPL 0, pres 1, def32 1, gran 1
 > processor eflags        = interrupt enabled, resume, IOPL = 0
 > current process        = 0 (mskc0 taskq)
 > Physical memory: 495 MB
 > Dumping 132 MB: 117 101 85 69 53 37 21 5
 > 
 > Reading symbols from /boot/kernel/linux.ko...Reading symbols from
 > /boot/kernel/linux.ko.symbols...done.
 > done.
 > Loaded symbols for /boot/kernel/linux.ko
 > #0  0xc08907a9 in doadump () at /usr/src/sys/kern/kern_shutdown.c:254
 > 254    }
 > (kgdb) bt
 > #0  0xc08907a9 in doadump () at /usr/src/sys/kern/kern_shutdown.c:254
 > #1  0xc04f7e37 in db_fncall (dummy1=-1067299898, dummy2=0,
 > dummy3=-718022488,
 >     dummy4=0xd533d898 "\200%t?") at /usr/src/sys/ddb/db_command.c:548
 > #2  0xc04f8214 in db_command (last_cmdp=0xc0da059c, cmd_table=0x0,
 > dopager=1)
 >     at /usr/src/sys/ddb/db_command.c:445
 > #3  0xc04f8352 in db_command_loop () at /usr/src/sys/ddb/db_command.c:498
 > #4  0xc04fa05e in db_trap (type=12, code=0) at
 > /usr/src/sys/ddb/db_main.c:229
 > #5  0xc08bf2d2 in kdb_reenter () at /usr/src/sys/kern/subr_kdb.c:398
 > #6  0xc0ba9b62 in trap_fatal (frame=0x1, eva=3735929098)
 >     at /usr/src/sys/i386/i386/trap.c:938
 > #7  0xc0baa483 in trap (frame=0xd533da80) at
 > /usr/src/sys/i386/i386/trap.c:339
 > #8  0xc0b8e4ab in Xlcall_syscall () at
 > /usr/src/sys/i386/i386/exception.s:241
 > #9  0xc0987410 in in_lltable_lookup (llt=0xc39e1000, flags=Variable
 > "flags" is not available.
 > )
 >     at /usr/src/sys/netinet/in.c:1380
 > #10 0xc0982470 in arpintr (m=0xc3baeb00) at
 > /usr/src/sys/netinet/if_ether.c:642
 > #11 0xc094227a in netisr_dispatch_src (proto=7, source=0, m=0xc0de)
 >     at /usr/src/sys/net/netisr.c:932
 > #12 0xc09424dd in netisr_unregister (nhp=0xc0de)
 >     at /usr/src/sys/net/netisr.c:583
 > #13 0xc093ac69 in ether_demux (ifp=0x0, m=0xc3baeb00)
 >     at /usr/src/sys/net/if_ethersubr.c:911
 > #14 0xc093b1ce in ether_output (ifp=0xc36ad400, m=0xc3baeb00,
 > dst=0xc0c55c27,
 >     ro=0x301010a) at /usr/src/sys/net/if_ethersubr.c:181
 > ---Type <return> to continue, or q <return> to quit---
 > #15 0xc070b032 in msk_handle_events (sc=0xc3686c00)
 >     at /usr/src/sys/dev/msk/if_msk.c:3048
 > #16 0xc070b828 in msk_int_task (arg=0xc3686c00, pending=1)
 >     at /usr/src/sys/dev/msk/if_msk.c:3625
 > #17 0xc08cac8c in taskqueue_run (queue=0xc36bf380)
 >     at /usr/src/sys/kern/subr_taskqueue.c:72
 > #18 0xc08cadcc in taskqueue_thread_loop (arg=0xc3686c8c)
 >     at /usr/src/sys/kern/subr_taskqueue.c:90
 > #19 0xc0869271 in fork_exit (callout=0xc08cad67 <taskqueue_thread_loop+64>,
 >     arg=0xc3686c8c, frame=0xd533dd38) at /usr/src/sys/kern/kern_fork.c:854
 > #20 0xc0b8e520 in Xatpic_intr0 () at atpic_vector.s:62
 > #21 0x00000000 in ?? ()
 > 
 
 I think it's not a bug of msk(4). Qin Li fixed the bug in arp code.
 See r198301.
 
 For watchdog timeout issues on 88E8053 controller, did you ever try
 disabling MSI? msk(4) was changed a lot since 7.0-RELEASE to
 support newer controllers and added several workarounds to address
 silicon bugs. So don't blindly apply experimental patches to your
 controller. 88E8053 also has a couple of hardware bugs but I guess
 msk(4) already incorporated required workarounds. So if you can
 reliably reproduce watchdog timeouts please let me know.