From: Robin Murphy <robin.murphy@arm.com>
To: Krishna Reddy <vdumpa@nvidia.com>
Cc: snikam@nvidia.com, thomasz@nvidia.com, jtukkinen@nvidia.com,
mperttunen@nvidia.com, praithatha@nvidia.com,
iommu@lists.linux-foundation.org, linux-kernel@vger.kernel.org,
talho@nvidia.com, yhsu@nvidia.com, linux-tegra@vger.kernel.org,
treding@nvidia.com, avanbrunt@nvidia.com,
linux-arm-kernel@lists.infradead.org
Subject: Re: [PATCH 3/7] iommu/arm-smmu: Add tlb_sync implementation hook
Date: Fri, 30 Aug 2019 16:23:13 +0100 [thread overview]
Message-ID: <554f8de1-1638-4eb9-59ae-8e1f0d786c44@arm.com> (raw)
In-Reply-To: <1567118827-26358-4-git-send-email-vdumpa@nvidia.com>
On 29/08/2019 23:47, Krishna Reddy wrote:
> tlb_sync hook allows nvidia smmu handle tlb sync
> across multiple SMMUs as necessary.
>
> Signed-off-by: Krishna Reddy <vdumpa@nvidia.com>
> ---
> drivers/iommu/arm-smmu-nvidia.c | 32 ++++++++++++++++++++++++++++++++
> drivers/iommu/arm-smmu.c | 8 +++++---
> drivers/iommu/arm-smmu.h | 4 ++++
> 3 files changed, 41 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/iommu/arm-smmu-nvidia.c b/drivers/iommu/arm-smmu-nvidia.c
> index d93ceda..a429b2c 100644
> --- a/drivers/iommu/arm-smmu-nvidia.c
> +++ b/drivers/iommu/arm-smmu-nvidia.c
> @@ -56,11 +56,43 @@ static void nsmmu_write_reg64(struct arm_smmu_device *smmu,
> writeq_relaxed(val, nsmmu_page(smmu, i, page) + offset);
> }
>
> +static void nsmmu_tlb_sync_wait(struct arm_smmu_device *smmu, int page,
> + int sync, int status, int inst)
> +{
> + u32 reg;
> + unsigned int spin_cnt, delay;
> +
> + for (delay = 1; delay < TLB_LOOP_TIMEOUT; delay *= 2) {
> + for (spin_cnt = TLB_SPIN_COUNT; spin_cnt > 0; spin_cnt--) {
> + reg = readl_relaxed(
> + nsmmu_page(smmu, inst, page) + status);
> + if (!(reg & sTLBGSTATUS_GSACTIVE))
> + return;
> + cpu_relax();
> + }
> + udelay(delay);
> + }
> + dev_err_ratelimited(smmu->dev,
> + "TLB sync timed out -- SMMU may be deadlocked\n");
> +}
> +
> +static void nsmmu_tlb_sync(struct arm_smmu_device *smmu, int page,
> + int sync, int status)
> +{
> + int i;
> +
> + arm_smmu_writel(smmu, page, sync, 0);
> +
> + for (i = 0; i < to_nsmmu(smmu)->num_inst; i++)
It might make more sense to make this the innermost loop, i.e.:
for (i = 0; i < nsmmu->num_inst; i++)
reg &= readl_relaxed(nsmmu_page(smmu, i, page)...
since polling the instances in parallel rather than in series seems like
it might be a bit more efficient.
> + nsmmu_tlb_sync_wait(smmu, page, sync, status, i);
> +}
> +
> static const struct arm_smmu_impl nsmmu_impl = {
> .read_reg = nsmmu_read_reg,
> .write_reg = nsmmu_write_reg,
> .read_reg64 = nsmmu_read_reg64,
> .write_reg64 = nsmmu_write_reg64,
> + .tlb_sync = nsmmu_tlb_sync,
> };
>
> struct arm_smmu_device *nvidia_smmu_impl_init(struct arm_smmu_device *smmu)
> diff --git a/drivers/iommu/arm-smmu.c b/drivers/iommu/arm-smmu.c
> index 46e1641..f5454e71 100644
> --- a/drivers/iommu/arm-smmu.c
> +++ b/drivers/iommu/arm-smmu.c
> @@ -52,9 +52,6 @@
> */
> #define QCOM_DUMMY_VAL -1
>
> -#define TLB_LOOP_TIMEOUT 1000000 /* 1s! */
> -#define TLB_SPIN_COUNT 10
> -
> #define MSI_IOVA_BASE 0x8000000
> #define MSI_IOVA_LENGTH 0x100000
>
> @@ -244,6 +241,11 @@ static void __arm_smmu_tlb_sync(struct arm_smmu_device *smmu, int page,
> unsigned int spin_cnt, delay;
> u32 reg;
>
> + if (smmu->impl->tlb_sync) {
> + smmu->impl->tlb_sync(smmu, page, sync, status);
What I'd hoped is that rather than needing a hook for this, you could
just override smmu_domain->tlb_ops from .init_context to wire up the
alternate .sync method directly. That would save this extra level of
indirection.
Robin.
> + return;
> + }
> +
> arm_smmu_writel(smmu, page, sync, QCOM_DUMMY_VAL);
> for (delay = 1; delay < TLB_LOOP_TIMEOUT; delay *= 2) {
> for (spin_cnt = TLB_SPIN_COUNT; spin_cnt > 0; spin_cnt--) {
> diff --git a/drivers/iommu/arm-smmu.h b/drivers/iommu/arm-smmu.h
> index 9645bf1..d3217f1 100644
> --- a/drivers/iommu/arm-smmu.h
> +++ b/drivers/iommu/arm-smmu.h
> @@ -207,6 +207,8 @@ enum arm_smmu_cbar_type {
> /* Maximum number of context banks per SMMU */
> #define ARM_SMMU_MAX_CBS 128
>
> +#define TLB_LOOP_TIMEOUT 1000000 /* 1s! */
> +#define TLB_SPIN_COUNT 10
>
> /* Shared driver definitions */
> enum arm_smmu_arch_version {
> @@ -336,6 +338,8 @@ struct arm_smmu_impl {
> int (*cfg_probe)(struct arm_smmu_device *smmu);
> int (*reset)(struct arm_smmu_device *smmu);
> int (*init_context)(struct arm_smmu_domain *smmu_domain);
> + void (*tlb_sync)(struct arm_smmu_device *smmu, int page, int sync,
> + int status);
> };
>
> static inline void __iomem *arm_smmu_page(struct arm_smmu_device *smmu, int n)
>
next prev parent reply other threads:[~2019-08-30 15:23 UTC|newest]
Thread overview: 33+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-08-29 22:47 [PATCH 0/7] Nvidia Arm SMMUv2 Implementation Krishna Reddy
2019-08-29 22:47 ` [PATCH 1/7] iommu/arm-smmu: add Nvidia SMMUv2 implementation Krishna Reddy
2019-08-30 15:02 ` Robin Murphy
2019-08-30 18:16 ` Krishna Reddy
2019-09-02 13:39 ` Robin Murphy
2019-09-03 1:07 ` Krishna Reddy
2019-08-29 22:47 ` [PATCH 2/7] dt-bindings: arm-smmu: Add binding for nvidia,smmu-v2 Krishna Reddy
2019-08-30 12:07 ` Mikko Perttunen
2019-08-30 15:13 ` Robin Murphy
2019-08-30 18:12 ` Krishna Reddy
2019-09-02 7:38 ` Thierry Reding
2019-08-29 22:47 ` [PATCH 3/7] iommu/arm-smmu: Add tlb_sync implementation hook Krishna Reddy
2019-08-30 11:14 ` Thierry Reding
2019-08-30 19:00 ` Krishna Reddy
2019-08-30 15:23 ` Robin Murphy [this message]
2019-08-30 18:05 ` Krishna Reddy
2019-08-30 22:49 ` Krishna Reddy
2019-09-02 13:00 ` Robin Murphy
2019-08-29 22:47 ` [PATCH 4/7] iommu/arm-smmu: Add global/context fault implementation hooks Krishna Reddy
2019-08-30 11:17 ` Thierry Reding
2019-08-30 19:16 ` Krishna Reddy
2019-08-30 15:43 ` Robin Murphy
2019-08-30 17:43 ` Krishna Reddy
2019-08-29 22:47 ` [PATCH 5/7] arm64: tegra: Add Memory controller DT node on T194 Krishna Reddy
2019-08-30 11:18 ` Thierry Reding
2019-08-29 22:47 ` [PATCH 6/7] arm64: tegra: Add DT node for T194 SMMU Krishna Reddy
2019-08-30 12:09 ` Mikko Perttunen
2019-08-30 18:39 ` Krishna Reddy
2019-08-30 15:44 ` Robin Murphy
2019-08-30 17:25 ` Krishna Reddy
2019-08-30 17:45 ` Robin Murphy
2019-08-30 18:35 ` Krishna Reddy
2019-08-29 22:47 ` [PATCH 7/7] arm64: tegra: enable SMMU for SDHCI and EQOS Krishna Reddy
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=554f8de1-1638-4eb9-59ae-8e1f0d786c44@arm.com \
--to=robin.murphy@arm.com \
--cc=avanbrunt@nvidia.com \
--cc=iommu@lists.linux-foundation.org \
--cc=jtukkinen@nvidia.com \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-tegra@vger.kernel.org \
--cc=mperttunen@nvidia.com \
--cc=praithatha@nvidia.com \
--cc=snikam@nvidia.com \
--cc=talho@nvidia.com \
--cc=thomasz@nvidia.com \
--cc=treding@nvidia.com \
--cc=vdumpa@nvidia.com \
--cc=yhsu@nvidia.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).