On Mon, Dec 15, 2014 at 6:48 PM, Paul Goyette <[email protected]> wrote: > On Mon, 15 Dec 2014, Ryota Ozaki wrote: > >> On Mon, Dec 15, 2014 at 3:38 PM, Ryota Ozaki <[email protected]> wrote: >>> >>> Hi, >>> >>> It seems ksyms (or something) is broken in a recent commit. >>> lockstat shows corrupted caller names like the following >>> output: >>> >>> ---- >>> $ sudo lockstat sleep 10 >>> Password: >>> Elapsed time: 9.99 seconds. >>> >>> -- Adaptive mutex spin >>> >>> Total% Count Time/ms Lock Caller >>> ------ ------- --------- ---------------------- >>> ------------------------------ >>> 95.76 35 0.22 <all> >>> 91.23 30 0.21 +103 >>> 3.41 2 0.01 +5f >>> 1.12 3 0.00 ?+71 >>> 4.24 2 0.01 fffffe827fba0f40 +1c >>> >>> -- Spin mutex spin >>> >>> Total% Count Time/ms Lock Caller >>> ------ ------- --------- ---------------------- >>> ------------------------------ >>> 95.16 18 0.03 fffffe827f732640 <all> >>> 92.41 17 0.02 fffffe827f732640 +95 >>> 2.76 1 0.00 fffffe827f732640 +6f >>> 2.72 1 0.00 fffffe827f7355c0 +103 >>> 2.12 1 0.00 fffffe827f7359c0 +103 >>> ---- >>> >>> I tested on a latest HEAD of -current kernel and userland >>> (self-build) and recent releng binaries built at 201412141850Z. >>> Both are amd64. >> >> >> The releng binaries at 201412131840Z works so the criminal >> is a commit between 201412131840Z and 201412141850Z :) > > > The only commit I see that appears as though it might be remotely associated > is: > > Module Name: src > Committed By: christos > Date: Sun Dec 14 18:14:15 UTC 2014 > > Modified Files: > src/sys/sys: ksyms.h > > Log Message: > Leave section 0 empty. Now gdb should recognize our dumped ksyms files. > > > > ------------------------------------------------------------------------- > | Paul Goyette | PGP Key fingerprint: | E-mail addresses: | > | (Retired) | FA29 0E3B 35AF E8AE 6651 | paul at whooppee.com | > | Network Engineer | 0786 F758 55DE 53BA 7731 | pgoyette at juniper.net | > | Kernel Developer | | pgoyette at netbsd.org | > -------------------------------------------------------------------------
Sure. And I confirmed a kernel and a lockstat binary of HEAD without the commit work correctly. Thanks, Paul! Christos, do you have a fix for the problem? or should I revert the commit? ozaki-r
