Skip to contentSkip to conversation

hw/arm/smmuv3: Fix incorrect reserved mask for SMMU CR0 register

PATCH In review [email protected] 3 participants 3 messages Jun 5, 2025 – Jun 9, 2025

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 hw/arm/smmuv3: Fix incorrect reserved mask for SMMU CR0 register

From: Tao Tang <[email protected]>

The current definition of the SMMU_CR0_RESERVED mask is incorrect. It mistakenly treats bit 10 (DPT_WALK_EN) as a reserved bit while treating bit 9 (RES0) as an implemented bit.

According to the SMMU architecture specification, the layout for CR0 is:

| 31:11| RES0           |
| 10   | DPT_WALK_EN    |
| 9    | RES0           |
| 8:6  | VMW            |
| 5    | RES0           |
| 4    | ATSCHK         |
| 3    | CMDQEN         |
| 2    | EVENTQEN       |
| 1    | PRIQEN         |
| 0    | SMMUEN         |
---
 hw/arm/smmuv3-internal.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
hw/arm/smmuv3-internal.h +1 -1
index b6b7399347..42ac77e654 100644
@@ -120,7 +120,7 @@ REG32(CR0,                 0x20)
     FIELD(CR0, EVENTQEN,      2, 1)
     FIELD(CR0, CMDQEN,        3, 1)
 
-#define SMMU_CR0_RESERVED 0xFFFFFC20
+#define SMMU_CR0_RESERVED 0xFFFFFA20
 
 REG32(CR0ACK,              0x24)
 REG32(CR1,                 0x28)
Signature · 1 line
2.34.1
Reviewed-by on v1 1/1
#1 lore raw

Hi,

Quoted 18 lines · On 6/5/25 4:26 AM, [email protected] wrote:

From: Tao Tang <[email protected]>

The current definition of the SMMU_CR0_RESERVED mask is incorrect. It mistakenly treats bit 10 (DPT_WALK_EN) as a reserved bit while treating bit 9 (RES0) as an implemented bit.

According to the SMMU architecture specification, the layout for CR0 is:

| 31:11| RES0           |
| 10   | DPT_WALK_EN    |
| 9    | RES0           |
| 8:6  | VMW            |
| 5    | RES0           |
| 4    | ATSCHK         |
| 3    | CMDQEN         |
| 2    | EVENTQEN       |
| 1    | PRIQEN         |
| 0    | SMMUEN         |
Quoted 1 line

Thanks!

Eric

Quoted 17 lines
---
 hw/arm/smmuv3-internal.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
hw/arm/smmuv3-internal.h +1 -1
index b6b7399347..42ac77e654 100644
@@ -120,7 +120,7 @@ REG32(CR0,                 0x20)
     FIELD(CR0, EVENTQEN,      2, 1)
     FIELD(CR0, CMDQEN,        3, 1)
 
-#define SMMU_CR0_RESERVED 0xFFFFFC20
+#define SMMU_CR0_RESERVED 0xFFFFFA20
 
 REG32(CR0ACK,              0x24)
 REG32(CR1,                 0x28)
#2 lore raw
In reply to [email protected]
Quoted 5 lines · On Thu, 5 Jun 2025 at 03:26, <[email protected]> wrote:

From: Tao Tang <[email protected]>

The current definition of the SMMU_CR0_RESERVED mask is incorrect. It mistakenly treats bit 10 (DPT_WALK_EN) as a reserved bit

This is because our implementation pre-dates the revision of the SMMUv3 spec which adds the DPT and this DPT_WALK_EN bit. So for us it is/was legitimately reserved. We use this constant value to update SMMU_CR0ACK when the guest writes to SMMU_CR0, so ideally we shouldn't mark bits as non-reserved unless we actually implement the feature that they correspond to.

We don't get this right for VMW and ATSCHK, which are both "RES0 unless the ID register indicates presence of some feature" fields, but which we allow the guest to write to (and cause updates to CR0ACK) anyway.

Quoted 2 lines

while
treating bit 9 (RES0) as an implemented bit.

This part is a bug; nice catch.

thanks
-- PMM