On 10/09/12 14:41, Christoffer Dall wrote: > On Mon, Sep 10, 2012 at 9:38 AM, Christoffer Dall > <c.dall@xxxxxxxxxxxxxxxxxxxxxx> wrote: >> On Mon, Sep 10, 2012 at 8:34 AM, Marc Zyngier <marc.zyngier@xxxxxxx> wrote: >>> When evincting pages from the host, we may end up in a situation >>> where a single page is getting evicted on multiple CPUs at the >>> same time. >>> >>> Because we serialize the evictions by holding pgd_lock, we make >>> sure that all CPUs but the first one will see the page as already >>> gone. If that page was the last in a section, the section is removed >>> as well. And all other CPUs will hit the BUG_ON condition, like this: >>> >>> [ 312.189234] ------------[ cut here ]------------ >>> [ 312.203056] kernel BUG at arch/arm/kvm/mmu.c:382! >>> [ 312.217134] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP THUMB2 >>> [ 312.235376] Modules linked in: >>> [ 312.244515] CPU: 0 Not tainted (3.6.0-rc3+ #40) >>> [ 312.259118] PC is at stage2_clear_pte+0x128/0x134 >>> [ 312.273193] LR is at kvm_unmap_hva+0x97/0xa0 >>> [ 312.285967] pc : [<c001e10c>] lr : [<c001ee0f>] psr: 60000133 >>> [ 312.285967] sp : caa25998 ip : df97a028 fp : 00800000 >>> [ 312.320355] r10: 873b5b5f r9 : c8654000 r8 : 01c55000 >>> [ 312.335990] r7 : 00000000 r6 : df249c00 r5 : c688fb80 r4 : df249ccc >>> [ 312.355532] r3 : 00000000 r2 : 2e001000 r1 : 00000000 r0 : 00000000 >>> [ 312.375076] Flags: nZCv IRQs on FIQs on Mode SVC_32 ISA Thumb Segment user >>> [ 312.396962] Control: 70c5387d Table: 8a9bbb00 DAC: fffffffd >>> [ 312.414161] Process hackbench (pid: 7207, stack limit = 0xcaa242f8) >>> >>> It is actually safe to remove the BUG_ONs and just return from the function >>> earlier. >>> >>> Signed-off-by: Marc Zyngier <marc.zyngier@xxxxxxx> >>> --- >>> arch/arm/kvm/mmu.c | 6 ++++-- >>> 1 file changed, 4 insertions(+), 2 deletions(-) >>> >>> diff --git a/arch/arm/kvm/mmu.c b/arch/arm/kvm/mmu.c >>> index e76c6a0..3610472 100644 >>> --- a/arch/arm/kvm/mmu.c >>> +++ b/arch/arm/kvm/mmu.c >>> @@ -376,10 +376,12 @@ static void stage2_clear_pte(struct kvm *kvm, phys_addr_t addr) >>> >>> pgd = kvm->arch.pgd + pgd_index(addr); >>> pud = pud_offset(pgd, addr); >>> - BUG_ON(pud_none(*pud)); >>> + if (pud_none(*pud)) >>> + return; /* Probably evicted already by another CPU */ >>> >>> pmd = pmd_offset(pud, addr); >>> - BUG_ON(pmd_none(*pmd)); >>> + if (pmd_none(*pmd)) >>> + return; /* Probably evicted already by another CPU */ >>> >>> pte = pte_offset_kernel(pmd, addr); >>> set_pte_ext(pte, __pte(0), 0); >>> -- >> >> > wait a second, I'm confused. We already have this fix, this was the > one reported by Andreas Sandberg, no? Duh. Indeed. I must have lost that patch during a rebase, and completely forgot about it. Time to move to a newer branch. Thanks, M. -- Jazz is not dead. It just smells funny... _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
