On 05/08/14 22:43, Philip Guenther wrote:
On Thu, May 8, 2014 at 2:59 PM, STeve Andre' <and...@msu.edu> wrote:
Twice now in three or so weeks, I've gotten a panic on my -current_amd64
W500 laptop. I've updated my tree several times during this time, and have
not seen other problems besides the known acpi heat problem.
Uh, what was the date of the cvs update of your kernel build when they
started? What was the cvs update date of your kernel before *that*? (I.e,
what's your best estimate of the window in which the change to the kernel
which triggered the panic occurred?
(What, you don't keep a log of the timestamps of your kernel
updates+builds? Doesn't everyone?)
Actually, I do keep past kernels so I have the build date for them.
I *thought* I had some notes on when this started but I am
ashamed to see that I didn't put them in a safe place.
I have both firefox and chrome running but I'm getting the feeling that
things
get more weird as I use lots of tabs in chrome.
You're pushing the vm subsystem enough to page. Since you have 8GB, I
wonder if you've raised yourkern.bufcachepercent, thus pushing on it harder.
Nope, I try to avoid the knobs when possible. It's been at 20%
ever since (bob?) raised it to 20%.
I don't think I'm swapping? At least I haven't seen top tell me that.
...In the past (like a year+) ago, there were times when chrome
went crazy with memory and I did swap. But chrome has gotten
better--I don't think I've seen it do that for some time now.
panic: softdep_deallocate_dependencies: dangling deps
Stopped at Debugger+0x5: leave
RUN AT LEAST 'trace' AND 'ps' AND INCLUDE OUTPUT WHEN REPORTING THIS PANIC!
IF RUNNING SMP, USE 'mach ddbcpu <#>' AND 'trace' ON OTHER PROCESSORS, TOO.
DO NOT EVEN BOTHER REPORTING THIS WITHOUT INCLUDING THAT INFORMATION!
ddb{1}> ddb{1}> Debugger() at Debugger+0x5
panic() at panic+0xfe
softdep_deallocate_dependencies() at softdep_deallocate_dependencies+0x1b
brelse() at brelse+0x61
ffs_write() at ffs_write+0x419
VOP_WRITE() at VOP_WRITE+0x3f
uvn_io() at uvn_io+0x1a0
uvm_pager_put() at uvm_pager_put+0x92
uvmpd_scan_inactive() at uvmpd_scan_inactive+0x5a3
uvmpd_scan() at uvmpd_scan+0x6e
uvm_pageout() at uvm_pageout+0x5b
This is *likely* to be a bug in the "don't cache pages being paged out"
commit, but it would be nice to be sure it didn't start before then...
Philip Guenther
I'm not sure how well I can pin this down. If I go too far back with
an older kernel I'll be out of sync with userland. Any suggestions
on how to test this more?
Thanks!
--STeve Andre'
