The following reply was made to PR kernel/6472; it has been noted by GNATS.
From: Cato Auestad <[email protected]> To: Joachim Schipper <[email protected]> Cc: [email protected] Subject: Re: kernel/6472: ThinkPad SL510 fails to boot (ACPI, acpitz related?) Date: Mon, 27 Sep 2010 05:04:37 +0200 On Sun, 2010-09-26 at 18:04 +0200, Joachim Schipper wrote: > >Number: 6472 > >Category: kernel > >Synopsis: ThinkPad SL510 fails to boot (ACPI, acpitz related?) > >Confidential: yes > >Severity: serious > >Priority: medium > >Responsible: bugs > >State: open > >Quarter: > >Keywords: > >Date-Required: > >Class: sw-bug > >Submitter-Id: unknown > >Arrival-Date: Sun Sep 26 16:20:01 GMT 2010 > >Closed-Date: > ast-Modified: > >Originator: > >Release: > >Organization: > >Environment: > System : OpenBSD 4.8 > Details : OpenBSD 4.8-current (GENERIC.MP) #2: Sun Sep 26 17:25:14 > CEST 2010 > > [email protected]:/usr/src/sys/arch/amd64/compile/GENERIC.MP > > Architecture: OpenBSD.amd64 > Machine : amd64 > >Description: > My Lenovo Thinkpad SL510 fails to boot if the acpitz driver is enabled. > > The attached patch skips the first few calls to mp_setperf() from > acpitz_refresh(). This makes the the machine boot reliably (on a kernel built > from the newest src), but is *very* stupid. > > >How-To-Repeat: > Boot the machine, it will always fail. > > With the attached patch and acpitz_skip_first_setperfs = 0 (i.e. with the > debugging information obtained from the patch below, but with no changes in > behaviour), the last thing on the console (after mtrr has attached) will be > > acpitz0: acpitz_cpu_setperf at <mp_setperf> called: acpitz_cpu_setperf(100) > > (the address rendered as "<mp_setperf>" was resolved via ddb.) > > Booting worked intermittently on older kernels (a couple months old at least > - > I've been very busy moving, and wanted to write an actually-useful bug > report.) > > Once booted with the use of the supplied patch, changing performance via > hw.setperf or apmd works fine, as does > > $ for i in `jot 100 1 100`; do sudo sysctl hw.setperf=100 & done > > The machine seems to boot reliably with acpitz_skip_first_setperfs = 6, and > occasionally with acpitz_skip_first_setperfs = 5. I've not observed any > succesful boots with acpitz_skip_first_setperfs = 4 and lower, but I didn't > try > very hard. > > >Fix: > The following patch seems to make the machine boot reliably (as well as > adding > some debug information). As noted above, it "is *very* stupid". > > Stupidity goes well with this machine, though: see > src/sys/arch/amd64/amd64/acpi_machdep.c r1.44 (someone misread the ACPI spec > and set the version number to 4 for a v4.00 machine - which is completely > wrong.) > > I'm happy to test a sane(r) patch. That seems odd. I installed -current on an SL510 which I bought less than a week ago, and it worked perfectly. If you have the exact same model (Type 2847) as I do and you can't boot it then it must be a recent change. Cato Auestad.
