Hi Julien
On 24.05.18 16:49, Julien Grall wrote:
Hi Artem,
Thank you for the report.
On 24/05/18 14:20, Artem Mygaiev wrote:
vgic_vcpu_pending_irq() uses find_next_bit() library function with
single 'unsigned long' variable, while it is designed to work with
memory regions. Nothing wrong is happening since 'offset' is set to 0
(other values could lead to memory corruption), but it would be more
correct to use the find_first_bit() function instead.
I don't understand the commit message. It is fine to use other offset
than 0 in find_next_bit as long as it is smaller than 32. There would be
no corruption happening.
Furthermore, find_first_bit(&apr, 32, 0) and find_next_bit(&apr, 32) are
equivalent because the former is just a macro using the latter (see
include/asm-arm/arm64/bitops.h).
So as it is the patch is not solving anything. However, I think this is
just a false positive. Coverity should be able to guess that it will not
go past the array (BITOP_WORD will turned into 0).
Absolutely agree with you. Probably my message was not clear enough -
with this particular patch I am not trying to fix a memory corruption,
there is no memory corruption in the code now. It is just the use of
functions: find_first_bit() is a better fit since the
vgic_vcpu_pending_irq() function does not need to go over memory region
and checks only one 32-bit variable. I have mentioned Coverity issue
here because this was a false positive detected after today's test run.
Coverity Scan issue 1469342
For future reference, please use the tag: "Coverity-ID: 1469342".
Thanks, will do.
Signed-off-by: Artem Mygaiev <artem_myga...@epam.com>
diff --git a/xen/arch/arm/gic-vgic.c b/xen/arch/arm/gic-vgic.c
index d831b35525..fd63906e9b 100644
--- a/xen/arch/arm/gic-vgic.c
+++ b/xen/arch/arm/gic-vgic.c
@@ -362,7 +362,7 @@ int vgic_vcpu_pending_irq(struct vcpu *v)
ASSERT(v == current);
mask_priority = gic_hw_ops->read_vmcr_priority();
- active_priority = find_next_bit(&apr, 32, 0);
+ active_priority = find_first_bit(&apr, 32);
spin_lock_irqsave(&v->arch.vgic.lock, flags);
Cheers,
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel