KVM: MMU: fix an IS_ERR() vs NULL bug
Subsystems (4)
PATCH 1/1 KVM: MMU: fix an IS_ERR() vs NULL bug
The alloc_workqueue() function does not return error pointers, it returns NULL on error. Update the check accordingly.
---
Obviously, I noticed that the patch says "propagate alloc_workqueue failure" so that's a puzzling thing. Merge issue perhaps? In linux-next it alloc_workqueue() returns NULL.
arch/x86/kvm/mmu/tdp_mmu.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
arch/x86/kvm/mmu/tdp_mmu.c +2 -2
@@ -22,8 +22,8 @@ int kvm_mmu_init_tdp_mmu(struct kvm *kvm)return 0 ; wq =alloc_workqueue ("kvm" ,WQ_UNBOUND |WQ_MEM_RECLAIM |WQ_CPU_INTENSIVE ,0 ); -if (IS_ERR (wq)) -return PTR_ERR (wq); +if (!wq) +return -ENOMEM ;/* This should not be changed for the lifetime of the VM. */ kvm->arch.tdp_mmu_enabled = true;
Signature · 1 line
2.20.1
Quoted 6 lines · Dan Carpenter
The
alloc_workqueue()function does not return error pointers, it returns NULL on error. Update the check accordingly.Fixes 1a3320dd2939 ("KVM: MMU: propagate alloc_workqueue failure")Signed-off-by Dan Carpenter <[email protected]>---
Quoted 3 lines
Obviously, I noticed that the patch says "propagate
alloc_workqueuefailure" so that's a puzzling thing. Merge issue perhaps? In linux-next italloc_workqueue()returns NULL.
No merge issue, just a goof. The "propagate" patch was added because KVM neglected to check for allocation failure, so at least it was a step in the right direction :-)
Quoted 9 lines
arch/x86/kvm/mmu/tdp_mmu.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)arch/x86/kvm/mmu/tdp_mmu.c +0
-0
arch/x86/kvm/mmu/tdp_mmu.c int kvm_mmu_init_tdp_mmu(struct kvm *kvm) · line 29
Quoted 10 lines
arch/x86/kvm/mmu/tdp_mmu.c +2
-2@@ -22,8 +22,8 @@ int kvm_mmu_init_tdp_mmu(struct kvm *kvm)return 0 ; wq =alloc_workqueue ("kvm" ,WQ_UNBOUND |WQ_MEM_RECLAIM |WQ_CPU_INTENSIVE ,0 ); -if (IS_ERR (wq)) -return PTR_ERR (wq); +if (!wq) +return -ENOMEM ;/* This should not be changed for the lifetime of the VM. */ kvm->arch.tdp_mmu_enabled = true;
Paolo, any objection to also returning '0' in all non-error paths? There's no need to return whether or not the TDP MMU is enabled since that's handled locally, and the "return 1" is rather odd.
arch/x86/kvm/mmu/mmu.c +1 -1
@@ -5779,7 +5779,7 @@ int kvm_mmu_init_vm(struct kvm *kvm)spin_lock_init (&kvm->arch.mmu_unsync_pages_lock); r =kvm_mmu_init_tdp_mmu (kvm); -if (r <0 ) +if (r)return r; node->track_write = kvm_mmu_pte_write;
arch/x86/kvm/mmu/tdp_mmu.c +1 -1
@@ -31,7 +31,7 @@ int kvm_mmu_init_tdp_mmu(struct kvm *kvm)spin_lock_init (&kvm->arch.tdp_mmu_pages_lock);INIT_LIST_HEAD (&kvm->arch.tdp_mmu_pages); kvm->arch.tdp_mmu_zap_wq = wq; -return 1 ; +return 0 ; }/* Arbitrarily returns true so that this may be used in if statements. */
Paolo Bonzini
maintainerQuoted 3 lines · Sean Christopherson
Paolo, any objection to also returning '0' in all non-error paths? There's no need to return whether or not the TDP MMU is enabled since that's handled locally, and the "return 1" is rather odd.
Well, I kept that because I thought there was a hidden reason for that.
At this point I was really about to press Enter and send out the pull
request, so...
Paolo