[PATCH 8/9] iommu/arm-smmu-v3: Directly encode CMDQ_OP_SYNC

[Mostafa Saleh]

On Fri, May 01, 2026 at 11:29:17AM -0300, Jason Gunthorpe wrote:
> Change the flow so the caller controls the CS field and remove the
> weird u64p_replace_bits() thing to override it.
> 
> Signed-off-by: Jason Gunthorpe <jgg at nvidia.com>

Reviewed-by: Mostafa Saleh <smostafa at google.com>

Thanks,
Mostafa

> ---
>  drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c | 42 ++++++++-------------
>  drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h | 16 ++++++--
>  2 files changed, 27 insertions(+), 31 deletions(-)
> 
> diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
> index 0cdf0752ff6d62..8147b9cdcc6b99 100644
> --- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
> +++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
> @@ -308,16 +308,6 @@ static int arm_smmu_cmdq_build_cmd(struct arm_smmu_cmd *cmd_out,
>  	case CMDQ_OP_TLBI_EL2_ASID:
>  		cmd[0] |= FIELD_PREP(CMDQ_TLBI_0_ASID, ent->tlbi.asid);
>  		break;
> -	case CMDQ_OP_CMD_SYNC:
> -		if (ent->sync.msiaddr) {
> -			cmd[0] |= FIELD_PREP(CMDQ_SYNC_0_CS, CMDQ_SYNC_0_CS_IRQ);
> -			cmd[1] |= ent->sync.msiaddr & CMDQ_SYNC_1_MSIADDR_MASK;
> -		} else {
> -			cmd[0] |= FIELD_PREP(CMDQ_SYNC_0_CS, CMDQ_SYNC_0_CS_SEV);
> -		}
> -		cmd[0] |= FIELD_PREP(CMDQ_SYNC_0_MSH, ARM_SMMU_SH_ISH);
> -		cmd[0] |= FIELD_PREP(CMDQ_SYNC_0_MSIATTR, ARM_SMMU_MEMATTR_OIWB);
> -		break;
>  	default:
>  		return -ENOENT;
>  	}
> @@ -350,23 +340,24 @@ static void arm_smmu_cmdq_build_sync_cmd(struct arm_smmu_cmd *cmd,
>  					 struct arm_smmu_cmdq *cmdq, u32 prod)
>  {
>  	struct arm_smmu_queue *q = &cmdq->q;
> -	struct arm_smmu_cmdq_ent ent = {
> -		.opcode = CMDQ_OP_CMD_SYNC,
> -	};
> +	u64 msiaddr = 0;
> +	unsigned int cs;
>  
>  	/*
>  	 * Beware that Hi16xx adds an extra 32 bits of goodness to its MSI
>  	 * payload, so the write will zero the entire command on that platform.
>  	 */
> -	if (smmu->options & ARM_SMMU_OPT_MSIPOLL) {
> -		ent.sync.msiaddr = q->base_dma + Q_IDX(&q->llq, prod) *
> -				   q->ent_dwords * 8;
> +	if (arm_smmu_cmdq_needs_busy_polling(smmu, cmdq)) {
> +		cs = CMDQ_SYNC_0_CS_NONE;
> +	} else if (smmu->options & ARM_SMMU_OPT_MSIPOLL) {
> +		cs = CMDQ_SYNC_0_CS_IRQ;
> +		msiaddr = q->base_dma + Q_IDX(&q->llq, prod) *
> +			  q->ent_dwords * 8;
> +	} else {
> +		cs = CMDQ_SYNC_0_CS_SEV;
>  	}
>  
> -	arm_smmu_cmdq_build_cmd(cmd, &ent);
> -	if (arm_smmu_cmdq_needs_busy_polling(smmu, cmdq))
> -		u64p_replace_bits(&cmd->data[0], CMDQ_SYNC_0_CS_NONE,
> -				  CMDQ_SYNC_0_CS);
> +	*cmd = arm_smmu_make_cmd_sync(cs, msiaddr);
>  }
>  
>  void __arm_smmu_cmdq_skip_err(struct arm_smmu_device *smmu,
> @@ -383,9 +374,6 @@ void __arm_smmu_cmdq_skip_err(struct arm_smmu_device *smmu,
>  	struct arm_smmu_cmd cmd;
>  	u32 cons = readl_relaxed(q->cons_reg);
>  	u32 idx = FIELD_GET(CMDQ_CONS_ERR, cons);
> -	struct arm_smmu_cmdq_ent cmd_sync = {
> -		.opcode = CMDQ_OP_CMD_SYNC,
> -	};
>  
>  	dev_err(smmu->dev, "CMDQ error (cons 0x%08x): %s\n", cons,
>  		idx < ARRAY_SIZE(cerror_str) ?  cerror_str[idx] : "Unknown");
> @@ -419,10 +407,10 @@ void __arm_smmu_cmdq_skip_err(struct arm_smmu_device *smmu,
>  		dev_err(smmu->dev, "\t0x%016llx\n", (unsigned long long)cmd.data[i]);
>  
>  	/* Convert the erroneous command into a CMD_SYNC */
> -	arm_smmu_cmdq_build_cmd(&cmd, &cmd_sync);
> -	if (arm_smmu_cmdq_needs_busy_polling(smmu, cmdq))
> -		u64p_replace_bits(&cmd.data[0], CMDQ_SYNC_0_CS_NONE,
> -				  CMDQ_SYNC_0_CS);
> +	cmd = arm_smmu_make_cmd_sync(
> +		arm_smmu_cmdq_needs_busy_polling(smmu, cmdq) ?
> +			CMDQ_SYNC_0_CS_NONE : CMDQ_SYNC_0_CS_SEV,
> +		0);
>  
>  	queue_write(Q_ENT(q, cons), cmd.data, q->ent_dwords);
>  }
> diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h
> index 194f73cabef5c9..538380de7d48a0 100644
> --- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h
> +++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h
> @@ -571,6 +571,18 @@ static inline struct arm_smmu_cmd arm_smmu_make_cmd_atc_inv_all(u32 sid,
>  	return arm_smmu_make_cmd_atc_inv(sid, ssid, 0, ATC_INV_SIZE_ALL);
>  }
>  
> +static inline struct arm_smmu_cmd arm_smmu_make_cmd_sync(unsigned int cs,
> +							 u64 msiaddr)
> +{
> +	struct arm_smmu_cmd cmd = arm_smmu_make_cmd_op(CMDQ_OP_CMD_SYNC);
> +
> +	cmd.data[0] |= FIELD_PREP(CMDQ_SYNC_0_CS, cs) |
> +		       FIELD_PREP(CMDQ_SYNC_0_MSH, ARM_SMMU_SH_ISH) |
> +		       FIELD_PREP(CMDQ_SYNC_0_MSIATTR, ARM_SMMU_MEMATTR_OIWB);
> +	cmd.data[1] |= msiaddr & CMDQ_SYNC_1_MSIADDR_MASK;
> +	return cmd;
> +}
> +
>  /* Event queue */
>  #define EVTQ_ENT_SZ_SHIFT		5
>  #define EVTQ_ENT_DWORDS			((1 << EVTQ_ENT_SZ_SHIFT) >> 3)
> @@ -648,10 +660,6 @@ struct arm_smmu_cmdq_ent {
>  			u8			tg;
>  			u64			addr;
>  		} tlbi;
> -
> -		struct {
> -			u64			msiaddr;
> -		} sync;
>  	};
>  };
>  
> -- 
> 2.43.0
>