The commit f36a74b9345a itself is good, but it causes a panic in a
Linux VM that runs on a Hyper-V host that doesn't have the 15-bit
Extended APIC ID support:
kernel BUG at arch/x86/kernel/apic/io_apic.c:2408!
This happens because the Hyper-V ioapic_ir_domain (which is defined in
drivers/iommu/hyperv-iommu.c) can not be found. Fix the panic by
properly claiming the only I/O APIC emulated by Hyper-V.
Cc: David Woodhouse <[email protected]>
Cc: Vitaly Kuznetsov <[email protected]>
Fixes: f36a74b9345a ("x86/ioapic: Use I/O-APIC ID for finding irqdomain, not index")
Signed-off-by: Dexuan Cui <[email protected]>
---
drivers/iommu/hyperv-iommu.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
This patch is for the tip.git tree's x86/apic branch.
diff --git a/drivers/iommu/hyperv-iommu.c b/drivers/iommu/hyperv-iommu.c
index 9438daa24fdb..1d21a0b5f724 100644
--- a/drivers/iommu/hyperv-iommu.c
+++ b/drivers/iommu/hyperv-iommu.c
@@ -105,8 +105,8 @@ static int hyperv_irq_remapping_select(struct irq_domain *d,
struct irq_fwspec *fwspec,
enum irq_domain_bus_token bus_token)
{
- /* Claim only the first (and only) I/OAPIC */
- return x86_fwspec_is_ioapic(fwspec) && fwspec->param[0] == 0;
+ /* Claim the only I/O APIC emulated by Hyper-V */
+ return x86_fwspec_is_ioapic(fwspec);
}
static const struct irq_domain_ops hyperv_ir_domain_ops = {
base-commit: d1adcfbb520c43c10fc22fcdccdd4204e014fb53
--
2.27.0
On Tue, 2020-12-01 at 16:45 -0800, Dexuan Cui wrote:
> The commit f36a74b9345a itself is good, but it causes a panic in a
> Linux VM that runs on a Hyper-V host that doesn't have the 15-bit
> Extended APIC ID support:
> kernel BUG at arch/x86/kernel/apic/io_apic.c:2408!
>
> This happens because the Hyper-V ioapic_ir_domain (which is defined in
> drivers/iommu/hyperv-iommu.c) can not be found. Fix the panic by
> properly claiming the only I/O APIC emulated by Hyper-V.
>
> Cc: David Woodhouse <[email protected]>
> Cc: Vitaly Kuznetsov <[email protected]>
> Fixes: f36a74b9345a ("x86/ioapic: Use I/O-APIC ID for finding irqdomain, not index")
> Signed-off-by: Dexuan Cui <[email protected]>
Oops, apologies for missing that.
Reviewed-by: David Woodhouse <[email protected]>
Amazon Development Centre (London) Ltd. Registered in England and Wales with registration number 04543232 with its registered office at 1 Principal Place, Worship Street, London EC2A 2FA, United Kingdom.
On Tue, Dec 01 2020 at 16:45, Dexuan Cui wrote:
> The commit f36a74b9345a itself is good, but it causes a panic in a
> Linux VM that runs on a Hyper-V host that doesn't have the 15-bit
> Extended APIC ID support:
> kernel BUG at arch/x86/kernel/apic/io_apic.c:2408!
This has nothing to do with the 15bit APIC ID support, really.
The point is that the select() function only matches when I/O APIC ID is
0, which is not guaranteed. That's independent of the 15bit extended
APIC ID feature. But the I/O-APIC ID is irrelevant on hyperv because
there is only one.
> This happens because the Hyper-V ioapic_ir_domain (which is defined in
> drivers/iommu/hyperv-iommu.c) can not be found. Fix the panic by
> properly claiming the only I/O APIC emulated by Hyper-V.
We don't fix a panic. We fix a bug in the code :)
I'll amend the changelog.
Thanks,
tglx
The following commit has been merged into the x86/apic branch of tip:
Commit-ID: 26ab12bb9d96133b7880141d68b5e01a8783de9d
Gitweb: https://git.kernel.org/tip/26ab12bb9d96133b7880141d68b5e01a8783de9d
Author: Dexuan Cui <[email protected]>
AuthorDate: Tue, 01 Dec 2020 16:45:10 -08:00
Committer: Thomas Gleixner <[email protected]>
CommitterDate: Wed, 02 Dec 2020 11:22:55 +01:00
iommu/hyper-v: Remove I/O-APIC ID check from hyperv_irq_remapping_select()
commit a491bb19f728 ("iommu/hyper-v: Implement select() method on remapping
irqdomain") restricted the irq_domain_ops::select() callback to match on
I/O-APIC index 0, which was correct until the parameter was changed to
carry the I/O APIC ID in commit f36a74b9345a.
If the ID is not 0 then the match fails. Therefore I/O-APIC init fails to
retrieve the parent irqdomain for the I/O-APIC resulting in a boot panic:
kernel BUG at arch/x86/kernel/apic/io_apic.c:2408!
Fix it by matching the I/O-APIC independent of the ID as there is only one
I/O APIC emulated by Hyper-V.
[ tglx: Amended changelog ]
Fixes: f36a74b9345a ("x86/ioapic: Use I/O-APIC ID for finding irqdomain, not index")
Signed-off-by: Dexuan Cui <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Reviewed-by: David Woodhouse <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
---
drivers/iommu/hyperv-iommu.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/iommu/hyperv-iommu.c b/drivers/iommu/hyperv-iommu.c
index 9438daa..1d21a0b 100644
--- a/drivers/iommu/hyperv-iommu.c
+++ b/drivers/iommu/hyperv-iommu.c
@@ -105,8 +105,8 @@ static int hyperv_irq_remapping_select(struct irq_domain *d,
struct irq_fwspec *fwspec,
enum irq_domain_bus_token bus_token)
{
- /* Claim only the first (and only) I/OAPIC */
- return x86_fwspec_is_ioapic(fwspec) && fwspec->param[0] == 0;
+ /* Claim the only I/O APIC emulated by Hyper-V */
+ return x86_fwspec_is_ioapic(fwspec);
}
static const struct irq_domain_ops hyperv_ir_domain_ops = {
> From: Thomas Gleixner <[email protected]>
> Sent: Wednesday, December 2, 2020 1:56 AM
>
> On Tue, Dec 01 2020 at 16:45, Dexuan Cui wrote:
> > The commit f36a74b9345a itself is good, but it causes a panic in a
> > Linux VM that runs on a Hyper-V host that doesn't have the 15-bit
> > Extended APIC ID support:
> > kernel BUG at arch/x86/kernel/apic/io_apic.c:2408!
>
> This has nothing to do with the 15bit APIC ID support, really.
>
> The point is that the select() function only matches when I/O APIC ID is
> 0, which is not guaranteed. That's independent of the 15bit extended
> APIC ID feature. But the I/O-APIC ID is irrelevant on hyperv because
> there is only one.
>
> > This happens because the Hyper-V ioapic_ir_domain (which is defined in
> > drivers/iommu/hyperv-iommu.c) can not be found. Fix the panic by
> > properly claiming the only I/O APIC emulated by Hyper-V.
>
> We don't fix a panic. We fix a bug in the code :)
>
> I'll amend the changelog.
>
> Thanks,
>
> tglx
Thank you for reworking the commit log, tglx!
Dexuan