Skip to contentSkip to conversation

iommu/io-pgtable-arm: Add support to use system cache

PATCH Superseded Vivek Gautam 2 participants 4 messages May 13, 2019 – May 15, 2019

Subsystems (3)

v4 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

PATCH v4 1/1 iommu/io-pgtable-arm: Add support to use system cache

Few Qualcomm platforms such as, sdm845 have an additional outer cache called as System cache, aka. Last level cache (LLC) that allows non-coherent devices to upgrade to using caching. This cache sits right before the DDR, and is tightly coupled with the memory controller. The clients using this cache request their slices from this system cache, make it active, and can then start using it.

There is a fundamental assumption that non-coherent devices can't access caches. This change adds an exception where they *can* use some level of cache despite still being non-coherent overall. The coherent devices that use cacheable memory, and CPU make use of this system cache by default.

Looking at memory types, we have following -
a) Normal uncached :- MAIR 0x44, inner non-cacheable,

                      outer non-cacheable;

b) Normal cached :- MAIR 0xff, inner read write-back non-transient,
outer read write-back non-transient; attribute setting for coherenet I/O devices.
and, for non-coherent i/o devices that can allocate in system cache another type gets added -
c) Normal sys-cached :- MAIR 0xf4, inner non-cacheable, outer read write-back non-transient

Coherent I/O devices use system cache by marking the memory as normal cached.
Non-coherent I/O devices should mark the memory as normal sys-cached in page tables to use system cache.

---

V3 version of this patch and related series can be found at [1].

This change is a realisation of following changes from downstream msm-4.9:
iommu: io-pgtable-arm: Implement IOMMU_USE_UPSTREAM_HINT[2]

Changes since v3:
- Dropping support to cache i/o page tables to system cache. Getting support for data buffers is the first step.
Removed io-pgtable quirk and related change to add domain attribute.

Glmark2 numbers on SDM845 based cheza board:

S.No.|	with LLC support   |	without LLC support
     |	for data buffers   |
---------------------------------------------------		
1    |	4480; 72.3fps      |	4042; 65.2fps
2    |	4500; 72.6fps      |	4039; 65.1fps
3    |	4523; 72.9fps	   |	4106; 66.2fps
4    |	4489; 72.4fps	   |	4104; 66.2fps
5    |	4518; 72.9fps	   |	4072; 65.7fps

[1] https://patchwork.kernel.org/cover/10772629/
[2] https://source.codeaurora.org/quic/la/kernel/msm-4.9/commit/?h=msm-4.9&id=d4c72c413ea27c43f60825193d4de9cb8ffd9602

 drivers/iommu/io-pgtable-arm.c | 9 ++++++++-
 include/linux/iommu.h          | 1 +
 2 files changed, 9 insertions(+), 1 deletion(-)
drivers/iommu/io-pgtable-arm.c +8 -1
index d3700ec15cbd..2dbafe697531 100644
@@ -167,10 +167,12 @@
 #define ARM_LPAE_MAIR_ATTR_MASK		0xff
 #define ARM_LPAE_MAIR_ATTR_DEVICE	0x04
 #define ARM_LPAE_MAIR_ATTR_NC		0x44
+#define ARM_LPAE_MAIR_ATTR_QCOM_SYS_CACHE	0xf4
 #define ARM_LPAE_MAIR_ATTR_WBRWA	0xff
 #define ARM_LPAE_MAIR_ATTR_IDX_NC	0
 #define ARM_LPAE_MAIR_ATTR_IDX_CACHE	1
 #define ARM_LPAE_MAIR_ATTR_IDX_DEV	2
+#define ARM_LPAE_MAIR_ATTR_IDX_QCOM_SYS_CACHE	3
 
 /* IOPTE accessors */
 #define iopte_deref(pte,d) __va(iopte_to_paddr(pte, d))
@@ -442,6 +444,9 @@ static arm_lpae_iopte arm_lpae_prot_to_pte(struct arm_lpae_io_pgtable *data,
 		else if (prot & IOMMU_CACHE)
 			pte |= (ARM_LPAE_MAIR_ATTR_IDX_CACHE
 				<< ARM_LPAE_PTE_ATTRINDX_SHIFT);
+		else if (prot & IOMMU_QCOM_SYS_CACHE)
+			pte |= (ARM_LPAE_MAIR_ATTR_IDX_QCOM_SYS_CACHE
+				<< ARM_LPAE_PTE_ATTRINDX_SHIFT);
 	} else {
 		pte = ARM_LPAE_PTE_HAP_FAULT;
 		if (prot & IOMMU_READ)
@@ -841,7 +846,9 @@ arm_64_lpae_alloc_pgtable_s1(struct io_pgtable_cfg *cfg, void *cookie)
 	      (ARM_LPAE_MAIR_ATTR_WBRWA
 	       << ARM_LPAE_MAIR_ATTR_SHIFT(ARM_LPAE_MAIR_ATTR_IDX_CACHE)) |
 	      (ARM_LPAE_MAIR_ATTR_DEVICE
-	       << ARM_LPAE_MAIR_ATTR_SHIFT(ARM_LPAE_MAIR_ATTR_IDX_DEV));
+	       << ARM_LPAE_MAIR_ATTR_SHIFT(ARM_LPAE_MAIR_ATTR_IDX_DEV)) |
+	      (ARM_LPAE_MAIR_ATTR_QCOM_SYS_CACHE
+	       << ARM_LPAE_MAIR_ATTR_SHIFT(ARM_LPAE_MAIR_ATTR_IDX_QCOM_SYS_CACHE));
 
 	cfg->arm_lpae_s1_cfg.mair[0] = reg;
 	cfg->arm_lpae_s1_cfg.mair[1] = 0;
include/linux/iommu.h +1 -0
index a815cf6f6f47..29dd2c624348 100644
@@ -31,6 +31,7 @@
 #define IOMMU_CACHE	(1 << 2) /* DMA cache coherency */
 #define IOMMU_NOEXEC	(1 << 3)
 #define IOMMU_MMIO	(1 << 4) /* e.g. things like MSI doorbells */
+#define IOMMU_QCOM_SYS_CACHE	(1 << 6)
 /*
  * Where the bus hardware includes a privilege level as part of its access type
  * markings, and certain devices are capable of issuing transactions marked as
Signature · 7 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation

_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu
reviewer
#1 lore raw
Quoted 66 lines · Vivek Gautam

Few Qualcomm platforms such as, sdm845 have an additional outer cache called as System cache, aka. Last level cache (LLC) that allows non-coherent devices to upgrade to using caching. This cache sits right before the DDR, and is tightly coupled with the memory controller. The clients using this cache request their slices from this system cache, make it active, and can then start using it.

There is a fundamental assumption that non-coherent devices can't access caches. This change adds an exception where they *can* use some level of cache despite still being non-coherent overall. The coherent devices that use cacheable memory, and CPU make use of this system cache by default.

Looking at memory types, we have following -
a) Normal uncached :- MAIR 0x44, inner non-cacheable,

                       outer non-cacheable;

b) Normal cached :- MAIR 0xff, inner read write-back non-transient,
outer read write-back non-transient; attribute setting for coherenet I/O devices.
and, for non-coherent i/o devices that can allocate in system cache another type gets added -
c) Normal sys-cached :- MAIR 0xf4, inner non-cacheable, outer read write-back non-transient

Coherent I/O devices use system cache by marking the memory as normal cached.
Non-coherent I/O devices should mark the memory as normal sys-cached in page tables to use system cache.

---

V3 version of this patch and related series can be found at [1].

This change is a realisation of following changes from downstream msm-4.9:
iommu: io-pgtable-arm: Implement IOMMU_USE_UPSTREAM_HINT[2]

Changes since v3:
- Dropping support to cache i/o page tables to system cache. Getting support for data buffers is the first step.
Removed io-pgtable quirk and related change to add domain attribute.

Glmark2 numbers on SDM845 based cheza board:

S.No.| with LLC support | without LLC support
| for data buffers |
---------------------------------------------------

1    |	4480; 72.3fps      |	4042; 65.2fps
2    |	4500; 72.6fps      |	4039; 65.1fps
3    |	4523; 72.9fps	   |	4106; 66.2fps
4    |	4489; 72.4fps	   |	4104; 66.2fps
5    |	4518; 72.9fps	   |	4072; 65.7fps

[1] https://patchwork.kernel.org/cover/10772629/
[2] https://source.codeaurora.org/quic/la/kernel/msm-4.9/commit/?h=msm-4.9&id=d4c72c413ea27c43f60825193d4de9cb8ffd9602

  drivers/iommu/io-pgtable-arm.c | 9 ++++++++-
  include/linux/iommu.h          | 1 +
  2 files changed, 9 insertions(+), 1 deletion(-)
drivers/iommu/io-pgtable-arm.c +0 -0
index d3700ec15cbd..2dbafe697531 100644

drivers/iommu/io-pgtable-arm.c line 175

Quoted 9 lines
drivers/iommu/io-pgtable-arm.c +2 -0
@@ -167,10 +167,12 @@
  #define ARM_LPAE_MAIR_ATTR_MASK		0xff
  #define ARM_LPAE_MAIR_ATTR_DEVICE	0x04
  #define ARM_LPAE_MAIR_ATTR_NC		0x44
+#define ARM_LPAE_MAIR_ATTR_QCOM_SYS_CACHE	0xf4
  #define ARM_LPAE_MAIR_ATTR_WBRWA	0xff
  #define ARM_LPAE_MAIR_ATTR_IDX_NC	0
  #define ARM_LPAE_MAIR_ATTR_IDX_CACHE	1
  #define ARM_LPAE_MAIR_ATTR_IDX_DEV	2
+#define ARM_LPAE_MAIR_ATTR_IDX_QCOM_SYS_CACHE	3

Here at the implementation level, I'd rather just call these what they are, i.e. s/QCOM_SYS_CACHE/INC_OWBRWA/.

drivers/iommu/io-pgtable-arm.c static arm_lpae_iopte arm_lpae_prot_to_pte(struct arm_lpae_io_pgtable *data, · line 447

Quoted 8 lines
drivers/iommu/io-pgtable-arm.c +1 -0
  /* IOPTE accessors */
  #define iopte_deref(pte,d) __va(iopte_to_paddr(pte, d))
@@ -442,6 +444,9 @@ static arm_lpae_iopte arm_lpae_prot_to_pte(struct arm_lpae_io_pgtable *data,
  		else if (prot & IOMMU_CACHE)
  			pte |= (ARM_LPAE_MAIR_ATTR_IDX_CACHE
  				<< ARM_LPAE_PTE_ATTRINDX_SHIFT);
+		else if (prot & IOMMU_QCOM_SYS_CACHE)

Where in the call stack is this going to be decided? (I don't recall the previous discussions ever really reaching a solid conclusion on how to separate responsibilities).

Quoted 21 lines
drivers/iommu/io-pgtable-arm.c +5 -1
+			pte |= (ARM_LPAE_MAIR_ATTR_IDX_QCOM_SYS_CACHE
+				<< ARM_LPAE_PTE_ATTRINDX_SHIFT);
  	} else {
  		pte = ARM_LPAE_PTE_HAP_FAULT;
  		if (prot & IOMMU_READ)
@@ -841,7 +846,9 @@ arm_64_lpae_alloc_pgtable_s1(struct io_pgtable_cfg *cfg, void *cookie)
  	      (ARM_LPAE_MAIR_ATTR_WBRWA
  	       << ARM_LPAE_MAIR_ATTR_SHIFT(ARM_LPAE_MAIR_ATTR_IDX_CACHE)) |
  	      (ARM_LPAE_MAIR_ATTR_DEVICE
-	       << ARM_LPAE_MAIR_ATTR_SHIFT(ARM_LPAE_MAIR_ATTR_IDX_DEV));
+	       << ARM_LPAE_MAIR_ATTR_SHIFT(ARM_LPAE_MAIR_ATTR_IDX_DEV)) |
+	      (ARM_LPAE_MAIR_ATTR_QCOM_SYS_CACHE
+	       << ARM_LPAE_MAIR_ATTR_SHIFT(ARM_LPAE_MAIR_ATTR_IDX_QCOM_SYS_CACHE));
 
  	cfg->arm_lpae_s1_cfg.mair[0] = reg;
  	cfg->arm_lpae_s1_cfg.mair[1] = 0;
include/linux/iommu.h +0 -0
index a815cf6f6f47..29dd2c624348 100644

include/linux/iommu.h line 34

Quoted 4 lines
include/linux/iommu.h +1 -0
@@ -31,6 +31,7 @@
  #define IOMMU_CACHE	(1 << 2) /* DMA cache coherency */
  #define IOMMU_NOEXEC	(1 << 3)
  #define IOMMU_MMIO	(1 << 4) /* e.g. things like MSI doorbells */
+#define IOMMU_QCOM_SYS_CACHE	(1 << 6)

Nit: 6 usually comes *after* 5 ;)

Plus although it's fairly self-evident that this value has *something* to do with Qcom system caches and isn't as generic as, say, IOMMU_PRIV, it probably still warrants some degree of comment.

Robin.

Quoted 4 lines
include/linux/iommu.h +0 -0
  /*
   * Where the bus hardware includes a privilege level as part of its access type
   * markings, and certain devices are capable of issuing transactions marked as
#2 lore raw

Hi Robin,

Quoted 72 lines · Robin Murphy
Quoted 70 lines · Vivek Gautam

Few Qualcomm platforms such as, sdm845 have an additional outer cache called as System cache, aka. Last level cache (LLC) that allows non-coherent devices to upgrade to using caching. This cache sits right before the DDR, and is tightly coupled with the memory controller. The clients using this cache request their slices from this system cache, make it active, and can then start using it.

There is a fundamental assumption that non-coherent devices can't access caches. This change adds an exception where they *can* use some level of cache despite still being non-coherent overall. The coherent devices that use cacheable memory, and CPU make use of this system cache by default.

Looking at memory types, we have following -
a) Normal uncached :- MAIR 0x44, inner non-cacheable,

                       outer non-cacheable;

b) Normal cached :- MAIR 0xff, inner read write-back non-transient,
outer read write-back non-transient; attribute setting for coherenet I/O devices.
and, for non-coherent i/o devices that can allocate in system cache another type gets added -
c) Normal sys-cached :- MAIR 0xf4, inner non-cacheable, outer read write-back non-transient

Coherent I/O devices use system cache by marking the memory as normal cached.
Non-coherent I/O devices should mark the memory as normal sys-cached in page tables to use system cache.

---

V3 version of this patch and related series can be found at [1].

This change is a realisation of following changes from downstream msm-4.9:
iommu: io-pgtable-arm: Implement IOMMU_USE_UPSTREAM_HINT[2]

Changes since v3:
- Dropping support to cache i/o page tables to system cache. Getting support for data buffers is the first step.
Removed io-pgtable quirk and related change to add domain attribute.

Glmark2 numbers on SDM845 based cheza board:

S.No.|        with LLC support   |    without LLC support
      |       for data buffers   |
---------------------------------------------------
1    |        4480; 72.3fps      |    4042; 65.2fps
2    |        4500; 72.6fps      |    4039; 65.1fps
3    |        4523; 72.9fps      |    4106; 66.2fps
4    |        4489; 72.4fps      |    4104; 66.2fps
5    |        4518; 72.9fps      |    4072; 65.7fps

[1] https://patchwork.kernel.org/cover/10772629/
[2] https://source.codeaurora.org/quic/la/kernel/msm-4.9/commit/?h=msm-4.9&id=d4c72c413ea27c43f60825193d4de9cb8ffd9602

  drivers/iommu/io-pgtable-arm.c | 9 ++++++++-
  include/linux/iommu.h          | 1 +
  2 files changed, 9 insertions(+), 1 deletion(-)
drivers/iommu/io-pgtable-arm.c +1 -0
index d3700ec15cbd..2dbafe697531 100644
@@ -167,10 +167,12 @@
  #define ARM_LPAE_MAIR_ATTR_MASK             0xff
  #define ARM_LPAE_MAIR_ATTR_DEVICE   0x04
  #define ARM_LPAE_MAIR_ATTR_NC               0x44
+#define ARM_LPAE_MAIR_ATTR_QCOM_SYS_CACHE    0xf4
Quoted 9 lines

drivers/iommu/io-pgtable-arm.c line 175

Quoted 5 lines
drivers/iommu/io-pgtable-arm.c +1 -0
  #define ARM_LPAE_MAIR_ATTR_WBRWA    0xff
  #define ARM_LPAE_MAIR_ATTR_IDX_NC   0
  #define ARM_LPAE_MAIR_ATTR_IDX_CACHE        1
  #define ARM_LPAE_MAIR_ATTR_IDX_DEV  2
+#define ARM_LPAE_MAIR_ATTR_IDX_QCOM_SYS_CACHE        3

Here at the implementation level, I'd rather just call these what they are, i.e. s/QCOM_SYS_CACHE/INC_OWBRWA/.

Thanks for the review.
Sure, will change this as suggested.

Quoted 13 lines

drivers/iommu/io-pgtable-arm.c static arm_lpae_iopte arm_lpae_prot_to_pte(struct arm_lpae_io_pgtable *data, · line 447

Quoted 8 lines
drivers/iommu/io-pgtable-arm.c +1 -0
  /* IOPTE accessors */
  #define iopte_deref(pte,d) __va(iopte_to_paddr(pte, d))
@@ -442,6 +444,9 @@ static arm_lpae_iopte arm_lpae_prot_to_pte(struct arm_lpae_io_pgtable *data,
              else if (prot & IOMMU_CACHE)
                      pte |= (ARM_LPAE_MAIR_ATTR_IDX_CACHE
                              << ARM_LPAE_PTE_ATTRINDX_SHIFT);
+             else if (prot & IOMMU_QCOM_SYS_CACHE)

Where in the call stack is this going to be decided? (I don't recall the previous discussions ever really reaching a solid conclusion on how to separate responsibilities).

Based on the last discussion [1], I understood that we may not want to expose these cache protections to DMA APIs. So such control would lie with the masters that are creating the individual domains. An example [2] of this is graphics on sdm845.
Please ignore the change in naming at [2] IOMMU_UPSTREAM_HINT in [2] is same as IOMMU_QCOM_SYS_CACHE here.

At that point [1] I also pointed to the fact that video that uses DMA APIs to handle
buffers too uses system cache on sdm845. In this case shouldn't we expose the protection controls to DMA APIs? Or would you suggest that such devices get iommu domains in the driver, and then update these protection flags?

[1] https://lkml.org/lkml/2018/12/4/790
[2] https://patchwork.kernel.org/patch/10302791/

Quoted 14 lines
Quoted 14 lines
drivers/iommu/io-pgtable-arm.c +5 -1
+                     pte |= (ARM_LPAE_MAIR_ATTR_IDX_QCOM_SYS_CACHE
+                             << ARM_LPAE_PTE_ATTRINDX_SHIFT);
      } else {
              pte = ARM_LPAE_PTE_HAP_FAULT;
              if (prot & IOMMU_READ)
@@ -841,7 +846,9 @@ arm_64_lpae_alloc_pgtable_s1(struct io_pgtable_cfg *cfg, void *cookie)
            (ARM_LPAE_MAIR_ATTR_WBRWA
             << ARM_LPAE_MAIR_ATTR_SHIFT(ARM_LPAE_MAIR_ATTR_IDX_CACHE)) |
            (ARM_LPAE_MAIR_ATTR_DEVICE
-            << ARM_LPAE_MAIR_ATTR_SHIFT(ARM_LPAE_MAIR_ATTR_IDX_DEV));
+            << ARM_LPAE_MAIR_ATTR_SHIFT(ARM_LPAE_MAIR_ATTR_IDX_DEV)) |
+           (ARM_LPAE_MAIR_ATTR_QCOM_SYS_CACHE
+            << ARM_LPAE_MAIR_ATTR_SHIFT(ARM_LPAE_MAIR_ATTR_IDX_QCOM_SYS_CACHE));
 
Quoted 13 lines

include/linux/iommu.h line 34

Quoted 11 lines
drivers/iommu/io-pgtable-arm.c +0 -0
      cfg->arm_lpae_s1_cfg.mair[0] = reg;
      cfg->arm_lpae_s1_cfg.mair[1] = 0;
include/linux/iommu.h +1 -0
index a815cf6f6f47..29dd2c624348 100644
@@ -31,6 +31,7 @@
  #define IOMMU_CACHE (1 << 2) /* DMA cache coherency */
  #define IOMMU_NOEXEC        (1 << 3)
  #define IOMMU_MMIO  (1 << 4) /* e.g. things like MSI doorbells */
+#define IOMMU_QCOM_SYS_CACHE (1 << 6)

Nit: 6 usually comes *after* 5 ;)

Sorry, pasting mistake.

Quoted 4 lines

Plus although it's fairly self-evident that this value has *something* to do with Qcom system caches and isn't as generic as, say, IOMMU_PRIV, it probably still warrants some degree of comment.

I will add the necessary comments.

Best regards
Vivek

Quoted 11 lines

Robin.

include/linux/iommu.h line 37

Quoted 4 lines
include/linux/iommu.h +0 -0
  /*
   * Where the bus hardware includes a privilege level as part of its access type
   * markings, and certain devices are capable of issuing transactions marked as

_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu

Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation
#3 lore raw
Quoted 68 lines · Vivek Gautam

Hi Robin,

Quoted 63 lines · Robin Murphy
Quoted 61 lines · Vivek Gautam

Few Qualcomm platforms such as, sdm845 have an additional outer cache called as System cache, aka. Last level cache (LLC) that allows non-coherent devices to upgrade to using caching. This cache sits right before the DDR, and is tightly coupled with the memory controller. The clients using this cache request their slices from this system cache, make it active, and can then start using it.

There is a fundamental assumption that non-coherent devices can't access caches. This change adds an exception where they *can* use some level of cache despite still being non-coherent overall. The coherent devices that use cacheable memory, and CPU make use of this system cache by default.

Looking at memory types, we have following -
a) Normal uncached :- MAIR 0x44, inner non-cacheable,

                       outer non-cacheable;

b) Normal cached :- MAIR 0xff, inner read write-back non-transient,
outer read write-back non-transient; attribute setting for coherenet I/O devices.
and, for non-coherent i/o devices that can allocate in system cache another type gets added -
c) Normal sys-cached :- MAIR 0xf4, inner non-cacheable, outer read write-back non-transient

Coherent I/O devices use system cache by marking the memory as normal cached.
Non-coherent I/O devices should mark the memory as normal sys-cached in page tables to use system cache.

---

V3 version of this patch and related series can be found at [1].

This change is a realisation of following changes from downstream msm-4.9:
iommu: io-pgtable-arm: Implement IOMMU_USE_UPSTREAM_HINT[2]

Changes since v3:
- Dropping support to cache i/o page tables to system cache. Getting support for data buffers is the first step.
Removed io-pgtable quirk and related change to add domain attribute.

Glmark2 numbers on SDM845 based cheza board:

S.No.|        with LLC support   |    without LLC support
      |       for data buffers   |
---------------------------------------------------
1    |        4480; 72.3fps      |    4042; 65.2fps
2    |        4500; 72.6fps      |    4039; 65.1fps
3    |        4523; 72.9fps      |    4106; 66.2fps
4    |        4489; 72.4fps      |    4104; 66.2fps
5    |        4518; 72.9fps      |    4072; 65.7fps

[1] https://patchwork.kernel.org/cover/10772629/
[2] https://source.codeaurora.org/quic/la/kernel/msm-4.9/commit/?h=msm-4.9&id=d4c72c413ea27c43f60825193d4de9cb8ffd9602

  drivers/iommu/io-pgtable-arm.c | 9 ++++++++-
  include/linux/iommu.h          | 1 +
  2 files changed, 9 insertions(+), 1 deletion(-)
Quoted 9 lines
Quoted 9 lines

drivers/iommu/io-pgtable-arm.c line 170

Quoted 9 lines
drivers/iommu/io-pgtable-arm.c +1 -0
index d3700ec15cbd..2dbafe697531 100644
@@ -167,10 +167,12 @@
  #define ARM_LPAE_MAIR_ATTR_MASK             0xff
  #define ARM_LPAE_MAIR_ATTR_DEVICE   0x04
  #define ARM_LPAE_MAIR_ATTR_NC               0x44
+#define ARM_LPAE_MAIR_ATTR_QCOM_SYS_CACHE    0xf4

s/QCOM_SYS_CACHE/INC_OWBRWA/ looks more appropriate here.

Quoted 8 lines
Quoted 8 lines

drivers/iommu/io-pgtable-arm.c line 175

Quoted 5 lines
drivers/iommu/io-pgtable-arm.c +1 -0
  #define ARM_LPAE_MAIR_ATTR_WBRWA    0xff
  #define ARM_LPAE_MAIR_ATTR_IDX_NC   0
  #define ARM_LPAE_MAIR_ATTR_IDX_CACHE        1
  #define ARM_LPAE_MAIR_ATTR_IDX_DEV  2
+#define ARM_LPAE_MAIR_ATTR_IDX_QCOM_SYS_CACHE        3

Here at the implementation level, I'd rather just call these what they are, i.e. s/QCOM_SYS_CACHE/INC_OWBRWA/.

Allow me to change this to - s/QCOM_SYS_CACHE/INC_OC, or s/QCOM_SYS_CACHE/INC_OCACHE to go inline with IDX_NC and IDX_CACHE. Sounds okay?

Quoted 91 lines
Quoted 1 line

Thanks for the review.
Sure, will change this as suggested.

Quoted 13 lines

drivers/iommu/io-pgtable-arm.c static arm_lpae_iopte arm_lpae_prot_to_pte(struct arm_lpae_io_pgtable *data, · line 447

Quoted 8 lines
drivers/iommu/io-pgtable-arm.c +1 -0
  /* IOPTE accessors */
  #define iopte_deref(pte,d) __va(iopte_to_paddr(pte, d))
@@ -442,6 +444,9 @@ static arm_lpae_iopte arm_lpae_prot_to_pte(struct arm_lpae_io_pgtable *data,
              else if (prot & IOMMU_CACHE)
                      pte |= (ARM_LPAE_MAIR_ATTR_IDX_CACHE
                              << ARM_LPAE_PTE_ATTRINDX_SHIFT);
+             else if (prot & IOMMU_QCOM_SYS_CACHE)

Where in the call stack is this going to be decided? (I don't recall the previous discussions ever really reaching a solid conclusion on how to separate responsibilities).

Based on the last discussion [1], I understood that we may not want to expose these cache protections to DMA APIs. So such control would lie with the masters that are creating the individual domains. An example [2] of this is graphics on sdm845.
Please ignore the change in naming at [2] IOMMU_UPSTREAM_HINT in [2] is same as IOMMU_QCOM_SYS_CACHE here.

At that point [1] I also pointed to the fact that video that uses DMA APIs to handle
buffers too uses system cache on sdm845. In this case shouldn't we expose the protection controls to DMA APIs? Or would you suggest that such devices get iommu domains in the driver, and then update these protection flags?

[1] https://lkml.org/lkml/2018/12/4/790
[2] https://patchwork.kernel.org/patch/10302791/

Quoted 14 lines
Quoted 14 lines
drivers/iommu/io-pgtable-arm.c +5 -1
+                     pte |= (ARM_LPAE_MAIR_ATTR_IDX_QCOM_SYS_CACHE
+                             << ARM_LPAE_PTE_ATTRINDX_SHIFT);
      } else {
              pte = ARM_LPAE_PTE_HAP_FAULT;
              if (prot & IOMMU_READ)
@@ -841,7 +846,9 @@ arm_64_lpae_alloc_pgtable_s1(struct io_pgtable_cfg *cfg, void *cookie)
            (ARM_LPAE_MAIR_ATTR_WBRWA
             << ARM_LPAE_MAIR_ATTR_SHIFT(ARM_LPAE_MAIR_ATTR_IDX_CACHE)) |
            (ARM_LPAE_MAIR_ATTR_DEVICE
-            << ARM_LPAE_MAIR_ATTR_SHIFT(ARM_LPAE_MAIR_ATTR_IDX_DEV));
+            << ARM_LPAE_MAIR_ATTR_SHIFT(ARM_LPAE_MAIR_ATTR_IDX_DEV)) |
+           (ARM_LPAE_MAIR_ATTR_QCOM_SYS_CACHE
+            << ARM_LPAE_MAIR_ATTR_SHIFT(ARM_LPAE_MAIR_ATTR_IDX_QCOM_SYS_CACHE));
 
Quoted 13 lines

include/linux/iommu.h line 34

Quoted 11 lines
drivers/iommu/io-pgtable-arm.c +0 -0
      cfg->arm_lpae_s1_cfg.mair[0] = reg;
      cfg->arm_lpae_s1_cfg.mair[1] = 0;
include/linux/iommu.h +1 -0
index a815cf6f6f47..29dd2c624348 100644
@@ -31,6 +31,7 @@
  #define IOMMU_CACHE (1 << 2) /* DMA cache coherency */
  #define IOMMU_NOEXEC        (1 << 3)
  #define IOMMU_MMIO  (1 << 4) /* e.g. things like MSI doorbells */
+#define IOMMU_QCOM_SYS_CACHE (1 << 6)

Nit: 6 usually comes *after* 5 ;)

Sorry, pasting mistake.

Quoted 4 lines

Plus although it's fairly self-evident that this value has *something* to do with Qcom system caches and isn't as generic as, say, IOMMU_PRIV, it probably still warrants some degree of comment.

I will add the necessary comments.

Best regards
Vivek

Quoted 11 lines

Robin.

include/linux/iommu.h line 37

Quoted 4 lines
include/linux/iommu.h +0 -0
  /*
   * Where the bus hardware includes a privilege level as part of its access type
   * markings, and certain devices are capable of issuing transactions marked as

_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu

--
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation

Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation