[PATCH v2 3/4] iommu/arm-smmu-v3: Use CMD_SYNC completion MSI

Robin Murphy robin.murphy at arm.com
Mon Oct 16 05:25:17 PDT 2017


On 13/10/17 19:32, Will Deacon wrote:
> Hi Robin,
> 
> This mostly looks good. Just a few comments below.
> 
> On Thu, Aug 31, 2017 at 02:44:27PM +0100, Robin Murphy wrote:
>> As an IRQ, the CMD_SYNC interrupt is not particularly useful, not least
>> because we often need to wait for sync completion within someone else's
>> IRQ handler anyway. However, when the SMMU is both coherent and supports
>> MSIs, we can have a lot more fun by not using it as an interrupt at all.
>> Following the example suggested in the architecture and using a write
>> targeting normal memory, we can let callers wait on a status variable
>> outside the lock instead of having to stall the entire queue or even
>> touch MMIO registers. Since multiple sync commands are guaranteed to
>> complete in order, a simple incrementing sequence count is all we need
>> to unambiguously support any realistic number of overlapping waiters.
>>
>> Signed-off-by: Robin Murphy <robin.murphy at arm.com>
>> ---
>>
>> v2: Remove redundant 'bool msi' command member, other cosmetic tweaks
>>
>>  drivers/iommu/arm-smmu-v3.c | 47 +++++++++++++++++++++++++++++++++++++++++++--
>>  1 file changed, 45 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c
>> index f066725298cd..311f482b93d5 100644
>> --- a/drivers/iommu/arm-smmu-v3.c
>> +++ b/drivers/iommu/arm-smmu-v3.c
>> @@ -377,7 +377,16 @@
>>  
>>  #define CMDQ_SYNC_0_CS_SHIFT		12
>>  #define CMDQ_SYNC_0_CS_NONE		(0UL << CMDQ_SYNC_0_CS_SHIFT)
>> +#define CMDQ_SYNC_0_CS_IRQ		(1UL << CMDQ_SYNC_0_CS_SHIFT)
>>  #define CMDQ_SYNC_0_CS_SEV		(2UL << CMDQ_SYNC_0_CS_SHIFT)
>> +#define CMDQ_SYNC_0_MSH_SHIFT		22
>> +#define CMDQ_SYNC_0_MSH_ISH		(3UL << CMDQ_SYNC_0_MSH_SHIFT)
>> +#define CMDQ_SYNC_0_MSIATTR_SHIFT	24
>> +#define CMDQ_SYNC_0_MSIATTR_OIWB	(0xfUL << CMDQ_SYNC_0_MSIATTR_SHIFT)
>> +#define CMDQ_SYNC_0_MSIDATA_SHIFT	32
>> +#define CMDQ_SYNC_0_MSIDATA_MASK	0xffffffffUL
>> +#define CMDQ_SYNC_1_MSIADDR_SHIFT	0
>> +#define CMDQ_SYNC_1_MSIADDR_MASK	0xffffffffffffcUL
>>  
>>  /* Event queue */
>>  #define EVTQ_ENT_DWORDS			4
>> @@ -409,6 +418,7 @@
>>  /* High-level queue structures */
>>  #define ARM_SMMU_POLL_TIMEOUT_US	100
>>  #define ARM_SMMU_CMDQ_DRAIN_TIMEOUT_US	1000000 /* 1s! */
>> +#define ARM_SMMU_SYNC_TIMEOUT_US	1000000 /* 1s! */
> 
> We only ever do this when waiting for the queue to drain, so may as well
> just reuse the drain timeout.

As you've discovered, we remove the "drain" case entirely in the end.

>>  #define MSI_IOVA_BASE			0x8000000
>>  #define MSI_IOVA_LENGTH			0x100000
>> @@ -504,6 +514,10 @@ struct arm_smmu_cmdq_ent {
>>  		} pri;
>>  
>>  		#define CMDQ_OP_CMD_SYNC	0x46
>> +		struct {
>> +			u32			msidata;
>> +			u64			msiaddr;
>> +		} sync;
>>  	};
>>  };
>>  
>> @@ -617,6 +631,9 @@ struct arm_smmu_device {
>>  	int				gerr_irq;
>>  	int				combined_irq;
>>  
>> +	atomic_t			sync_nr;
>> +	u32				sync_count;
> 
> It's probably worth sticking these in separate cachelines so we don't
> get spurious wakeups when sync_nr is incremented. (yes, I know it should
> be the ERG, but that can be unreasonably huge!).

Good point - we've got 8K of bitmaps embedded in the structure anyway,
so even maximum ERG separation is easily possible.

>> +
>>  	unsigned long			ias; /* IPA */
>>  	unsigned long			oas; /* PA */
>>  	unsigned long			pgsize_bitmap;
>> @@ -878,7 +895,13 @@ static int arm_smmu_cmdq_build_cmd(u64 *cmd, struct arm_smmu_cmdq_ent *ent)
>>  		}
>>  		break;
>>  	case CMDQ_OP_CMD_SYNC:
>> -		cmd[0] |= CMDQ_SYNC_0_CS_SEV;
>> +		if (ent->sync.msiaddr)
>> +			cmd[0] |= CMDQ_SYNC_0_CS_IRQ;
>> +		else
>> +			cmd[0] |= CMDQ_SYNC_0_CS_SEV;
>> +		cmd[0] |= CMDQ_SYNC_0_MSH_ISH | CMDQ_SYNC_0_MSIATTR_OIWB;
>> +		cmd[0] |= (u64)ent->sync.msidata << CMDQ_SYNC_0_MSIDATA_SHIFT;
>> +		cmd[1] |= ent->sync.msiaddr & CMDQ_SYNC_1_MSIADDR_MASK;
>>  		break;
>>  	default:
>>  		return -ENOENT;
>> @@ -964,21 +987,40 @@ static void arm_smmu_cmdq_issue_cmd(struct arm_smmu_device *smmu,
>>  	spin_unlock_irqrestore(&smmu->cmdq.lock, flags);
>>  }
>>  
>> +static int arm_smmu_sync_poll_msi(struct arm_smmu_device *smmu, u32 sync_idx)
>> +{
>> +	ktime_t timeout = ktime_add_us(ktime_get(), ARM_SMMU_SYNC_TIMEOUT_US);
>> +	u32 val = smp_cond_load_acquire(&smmu->sync_count,
>> +					(int)(VAL - sync_idx) >= 0 ||
>> +					!ktime_before(ktime_get(), timeout));
>> +
>> +	return (int)(val - sync_idx) < 0 ? -ETIMEDOUT : 0;
> 
> There are some theoretical overflow issues here which I don't think will
> ever occur in practice, but deserve at least a comment to explain why.

Even if we did have 2^31 or more CPUs, the size of a queue is bounded at
2^20, so we can never have enough in-flight syncs to get near to an
overflow problem. I can certainly document that if you like.

>> +}
>> +
>>  static void arm_smmu_cmdq_issue_sync(struct arm_smmu_device *smmu)
>>  {
>>  	u64 cmd[CMDQ_ENT_DWORDS];
>>  	unsigned long flags;
>>  	bool wfe = !!(smmu->features & ARM_SMMU_FEAT_SEV);
>> +	bool msi = (smmu->features & ARM_SMMU_FEAT_MSI) &&
>> +		   (smmu->features & ARM_SMMU_FEAT_COHERENCY);
> 
> I don't think this is sufficient for the case where we fail to setup MSIs
> and fall back on legacy IRQs.

Remember this 'MSI' is going nowhere near the GIC, so the IRQ
configuration is irrelevant (especially after patch #2) - the feature
bits tell us "is the SMMU capable of generating sync-completion writes?"
and "are those writes coherent?", which is precisely what matters here.

>>  	struct arm_smmu_cmdq_ent ent = { .opcode = CMDQ_OP_CMD_SYNC };
>>  	int ret;
>>  
>> +	if (msi) {
>> +		ent.sync.msidata = atomic_inc_return(&smmu->sync_nr);
> 
> I don't think you need barrier semantics here.

You mean atomic_inc_return_relaxed() would be sufficient? TBH I don't
think I'd given this any thought - I guess the coherency protocols make
it impossible to do an atomic op on stale data, so that seems reasonable.
>> +		ent.sync.msiaddr = virt_to_phys(&smmu->sync_count);
>> +	}
>>  	arm_smmu_cmdq_build_cmd(cmd, &ent);
>>  
>>  	spin_lock_irqsave(&smmu->cmdq.lock, flags);
>>  	arm_smmu_cmdq_insert_cmd(smmu, cmd);
>> -	ret = queue_poll_cons(&smmu->cmdq.q, true, wfe);
>> +	if (!msi)
>> +		ret = queue_poll_cons(&smmu->cmdq.q, true, wfe);
>>  	spin_unlock_irqrestore(&smmu->cmdq.lock, flags);
>>  
>> +	if (msi)
>> +		ret = arm_smmu_sync_poll_msi(smmu, ent.sync.msidata);
> 
> This looks like the queue polling should be wrapped up in a function that
> returns with the spinlock released.

I did ponder that, but I can't help finding such asymmetric interfaces
pretty grim, and things do get better again once both cases can wait
outside the lock.

Robin.



More information about the linux-arm-kernel mailing list