2023-03-15 18:15:16

by D Scott Phillips

[permalink] [raw]
Subject: [PATCH RESEND] xhci: also avoid the XHCI_ZERO_64B_REGS quirk with a passthrough iommu

Previously the quirk was skipped when no iommu was present. The same
rationale for skipping the quirk also applies in the iommu.passthrough=1
case.

Skip applying the XHCI_ZERO_64B_REGS quirk if the device's iommu domain is
passthrough.

Fixes: 12de0a35c996 ("xhci: Add quirk to zero 64bit registers on Renesas PCIe controllers")
Signed-off-by: D Scott Phillips <[email protected]>
---
drivers/usb/host/xhci.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/drivers/usb/host/xhci.c b/drivers/usb/host/xhci.c
index 6183ce8574b1..bdb6dd819a3b 100644
--- a/drivers/usb/host/xhci.c
+++ b/drivers/usb/host/xhci.c
@@ -9,6 +9,7 @@
*/

#include <linux/pci.h>
+#include <linux/iommu.h>
#include <linux/iopoll.h>
#include <linux/irq.h>
#include <linux/log2.h>
@@ -228,6 +229,7 @@ int xhci_reset(struct xhci_hcd *xhci, u64 timeout_us)
static void xhci_zero_64b_regs(struct xhci_hcd *xhci)
{
struct device *dev = xhci_to_hcd(xhci)->self.sysdev;
+ struct iommu_domain *domain;
int err, i;
u64 val;
u32 intrs;
@@ -246,7 +248,9 @@ static void xhci_zero_64b_regs(struct xhci_hcd *xhci)
* an iommu. Doing anything when there is no iommu is definitely
* unsafe...
*/
- if (!(xhci->quirks & XHCI_ZERO_64B_REGS) || !device_iommu_mapped(dev))
+ domain = iommu_get_domain_for_dev(dev);
+ if (!(xhci->quirks & XHCI_ZERO_64B_REGS) || !domain ||
+ domain->type == IOMMU_DOMAIN_IDENTITY)
return;

xhci_info(xhci, "Zeroing 64bit base registers, expecting fault\n");
--
2.39.1



2023-03-16 09:12:06

by Marc Zyngier

[permalink] [raw]
Subject: Re: [PATCH RESEND] xhci: also avoid the XHCI_ZERO_64B_REGS quirk with a passthrough iommu

On Wed, 15 Mar 2023 18:14:40 +0000,
D Scott Phillips <[email protected]> wrote:
>
> Previously the quirk was skipped when no iommu was present. The same
> rationale for skipping the quirk also applies in the iommu.passthrough=1
> case.
>
> Skip applying the XHCI_ZERO_64B_REGS quirk if the device's iommu domain is
> passthrough.
>
> Fixes: 12de0a35c996 ("xhci: Add quirk to zero 64bit registers on Renesas PCIe controllers")
> Signed-off-by: D Scott Phillips <[email protected]>

Acked-by: Marc Zyngier <[email protected]>

M.

--
Without deviation from the norm, progress is not possible.

2023-03-17 15:17:55

by Mathias Nyman

[permalink] [raw]
Subject: Re: [PATCH RESEND] xhci: also avoid the XHCI_ZERO_64B_REGS quirk with a passthrough iommu

On 16.3.2023 11.11, Marc Zyngier wrote:
> On Wed, 15 Mar 2023 18:14:40 +0000,
> D Scott Phillips <[email protected]> wrote:
>>
>> Previously the quirk was skipped when no iommu was present. The same
>> rationale for skipping the quirk also applies in the iommu.passthrough=1
>> case.
>>
>> Skip applying the XHCI_ZERO_64B_REGS quirk if the device's iommu domain is
>> passthrough.
>>
>> Fixes: 12de0a35c996 ("xhci: Add quirk to zero 64bit registers on Renesas PCIe controllers")
>> Signed-off-by: D Scott Phillips <[email protected]>
>
> Acked-by: Marc Zyngier <[email protected]>
>
> M.

Thanks, adding to queue

-Mathias