On (08/12/09 20:18), Len Brown wrote: > > Kernel oopses on early boot stages (boots normally with boot_delay=10). > > > rc5-git8 > > Please find which kernel this failure starts in. > Start by testing 2.6.30.stable so we can tell > if this is a regression in the current RC. > OK... I'll try to bisect. It'll take time. I consider 7 successfull boots to mark that kernel is ok (it may be wrong). -git8 maxcpus=1 + + + + + + + acpi_serialize=1 + - boot_delay=1 + + + + + + + boot_delay=0 + + - vga=791 + + + + + + + > What is the difference between the two dmesg? > they have the same cmdline, neither has boot_delay, > and neither dmesg shows the oops. > boot params are the same. Both kernels booted successfully. There are no oopses in syslog (I guess 0.21 is too early for logd to start). The only difference (which I find suspicious) is around 0.21 second: [ 0.205668] ACPI Error (dswload-0790): [PCFG] Namespace lookup failure, AE_ALREADY_EXISTS [ 0.205899] ACPI Exception: AE_ALREADY_EXISTS, During name lookup/catalog 20090521 psloop-227 [ 0.206124] ACPI Error (psparse-0537): Method parse/execution failed [\_SB_.RBPE] (Node f7017678), AE_ALREADY_EXISTS [ 0.206417] ACPI: Marking method RBPE as Serialized because of AE_ALREADY_EXISTS error [ 0.206773] ACPI Error (psparse-0537): Method parse/execution failed [\_SB_.PCI0.P0P4.XCRD._STA] (Node f7015648), AE_ALREADY_EXISTS [ 0.207070] ACPI: Marking method _STA as Serialized because of AE_ALREADY_EXISTS error [ 0.207326] ACPI Error (uteval-0256): Method execution failed [\_SB_.PCI0.P0P4.XCRD._STA] (Node f7015648), AE_ALREADY_EXISTS [ 0.213503] ACPI Error: Could not map memory at 0000000000000000, size 0 20090521 exregion-173 [ 0.213734] ACPI Exception: AE_NO_MEMORY, Returned by Handler for [SystemMemory] 20090521 evregion-424 [ 0.213963] ACPI Error (psparse-0537): Method parse/execution failed [\_SB_.RBPE] (Node f7017678), AE_NO_MEMORY [ 0.214286] ACPI Error (psparse-0537): Method parse/execution failed [\_GPE._L01] (Node f701d198), AE_NO_MEMORY [ 0.214616] ACPI Exception: AE_NO_MEMORY, while evaluating GPE method [_L01] 20090521 evgpe-568 > The fact that boot_delay=10 makes the problem go away > suggests some sort of race condition. I think it is. > Is 10 the smallest value of printk_time that works, or is it a random value? > No. 1 is the smallest. > Do you still see the oops with CONFIG_PRINTK_TIME=n? > Yes. > Do you still see the oops with maxcpus=1? > No. > Do you still see the oops with "acpi_serialize=1"? > Yes. Sergey
Attachment:
signature.asc
Description: Digital signature
