Skip to contentSkip to conversation

Revert "dma-contiguous: do not allocate a single page from CMA area"

PATCH In review Nicolin Chen 5 participants 6 messages Feb 26, 2019 – Feb 27, 2019

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Revert "dma-contiguous: do not allocate a single page from CMA area"

This reverts commit d222e42e88168fd67e6d131984b86477af1fc256.

The original change breaks omap dss:
omapdss_dispc 58001000.dispc:
dispc_errata_i734_wa_init: dma_alloc_writecombine failed

Let's revert it first and then find a safer solution instead.

---
Tony,

Would you please test and verify? Thanks!

 kernel/dma/contiguous.c | 22 +++-------------------
 1 file changed, 3 insertions(+), 19 deletions(-)
kernel/dma/contiguous.c +3 -19
index 09074bd04793..b2a87905846d 100644
@@ -186,32 +186,16 @@ int __init dma_contiguous_reserve_area(phys_addr_t size, phys_addr_t base,
  *
  * This function allocates memory buffer for specified device. It uses
  * device specific contiguous memory area if available or the default
- * global one.
- *
- * However, it skips one-page size of allocations from the global area.
- * As the addresses within one page are always contiguous, so there is
- * no need to waste CMA pages for that kind; it also helps reduce the
- * fragmentations in the CMA area. So a caller should be the rebounder
- * in such case to allocate a normal page upon NULL return value.
- *
- * Requires architecture specific dev_get_cma_area() helper function.
+ * global one. Requires architecture specific dev_get_cma_area() helper
+ * function.
  */
 struct page *dma_alloc_from_contiguous(struct device *dev, size_t count,
 				       unsigned int align, bool no_warn)
 {
-	struct cma *cma;
-
 	if (align > CONFIG_CMA_ALIGNMENT)
 		align = CONFIG_CMA_ALIGNMENT;
 
-	if (dev && dev->cma_area)
-		cma = dev->cma_area;
-	else if (count > 1)
-		cma = dma_contiguous_default_area;
-	else
-		return NULL;
-
-	return cma_alloc(cma, count, align, no_warn);
+	return cma_alloc(dev_get_cma_area(dev), count, align, no_warn);
 }
 
 /**
Signature · 1 line
2.17.1
reviewer
#1 lore raw
Quoted 7 lines · Nicolin Chen

This reverts commit d222e42e88168fd67e6d131984b86477af1fc256.

The original change breaks omap dss:
omapdss_dispc 58001000.dispc:
dispc_errata_i734_wa_init: dma_alloc_writecombine failed

Let's revert it first and then find a safer solution instead.

Ah, I think I see the problem - once arch/arm's __dma_alloc() has decided to use CMA (because dev_get_cma_area(dev) returns the global area), it then won't fall back to trying a regular page allocation if dma_alloc_from_contiguous() returns NULL. Thus anything on 32-bit Arm trying to allocate a single-page buffer in blockable context with a CMA-enabled config is just going to fail. Similarly, it looks like none of the DMA_ATTR_FORCE_CONTIGUOUS cases are prepared to handle this change either (amd_iommu appears technically affected, but is already using dma_alloc_from_contiguous() backwards compared to everyone else, hmm).

I guess the question is whether to add alloc_page()/free_page() fallbacks to those call sites, or stuff them directly into the CMA helpers here.

Robin.

Quoted 51 lines

---
Tony,

Would you please test and verify? Thanks!

  kernel/dma/contiguous.c | 22 +++-------------------
  1 file changed, 3 insertions(+), 19 deletions(-)
kernel/dma/contiguous.c +3 -19
index 09074bd04793..b2a87905846d 100644
@@ -186,32 +186,16 @@ int __init dma_contiguous_reserve_area(phys_addr_t size, phys_addr_t base,
   *
   * This function allocates memory buffer for specified device. It uses
   * device specific contiguous memory area if available or the default
- * global one.
- *
- * However, it skips one-page size of allocations from the global area.
- * As the addresses within one page are always contiguous, so there is
- * no need to waste CMA pages for that kind; it also helps reduce the
- * fragmentations in the CMA area. So a caller should be the rebounder
- * in such case to allocate a normal page upon NULL return value.
- *
- * Requires architecture specific dev_get_cma_area() helper function.
+ * global one. Requires architecture specific dev_get_cma_area() helper
+ * function.
   */
  struct page *dma_alloc_from_contiguous(struct device *dev, size_t count,
  				       unsigned int align, bool no_warn)
  {
-	struct cma *cma;
-
  	if (align > CONFIG_CMA_ALIGNMENT)
  		align = CONFIG_CMA_ALIGNMENT;
 
-	if (dev && dev->cma_area)
-		cma = dev->cma_area;
-	else if (count > 1)
-		cma = dma_contiguous_default_area;
-	else
-		return NULL;
-
-	return cma_alloc(cma, count, align, no_warn);
+	return cma_alloc(dev_get_cma_area(dev), count, align, no_warn);
  }
 
  /**
Tested-by
#2 lore raw
Quoted 8 lines · Robin Murphy
Quoted 7 lines · Nicolin Chen

This reverts commit d222e42e88168fd67e6d131984b86477af1fc256.

The original change breaks omap dss:
omapdss_dispc 58001000.dispc:
dispc_errata_i734_wa_init: dma_alloc_writecombine failed

Let's revert it first and then find a safer solution instead.

Sounds like a good idea since we're only have few days left before the merge window.

Quoted 10 lines

Ah, I think I see the problem - once arch/arm's __dma_alloc() has decided to use CMA (because dev_get_cma_area(dev) returns the global area), it then won't fall back to trying a regular page allocation if dma_alloc_from_contiguous() returns NULL. Thus anything on 32-bit Arm trying to allocate a single-page buffer in blockable context with a CMA-enabled config is just going to fail. Similarly, it looks like none of the DMA_ATTR_FORCE_CONTIGUOUS cases are prepared to handle this change either (amd_iommu appears technically affected, but is already using dma_alloc_from_contiguous() backwards compared to everyone else, hmm).

Quoted 2 lines

I guess the question is whether to add alloc_page()/free_page() fallbacks to those call sites, or stuff them directly into the CMA helpers here.

Well if you come up with some test patch, I can easily test it :)

Quoted 1 line
Quoted 1 line

Would you please test and verify? Thanks!

Yes this revert works for me:

#3 lore raw
In reply to Robin Murphy
Quoted 9 lines · Robin Murphy
Quoted 7 lines · Nicolin Chen

This reverts commit d222e42e88168fd67e6d131984b86477af1fc256.

The original change breaks omap dss:
omapdss_dispc 58001000.dispc:
dispc_errata_i734_wa_init: dma_alloc_writecombine failed

Let's revert it first and then find a safer solution instead.

Quoted 9 lines

Ah, I think I see the problem - once arch/arm's __dma_alloc() has decided to use CMA (because dev_get_cma_area(dev) returns the global area), it then won't fall back to trying a regular page allocation if dma_alloc_from_contiguous() returns NULL. Thus anything on 32-bit Arm trying to allocate a single-page buffer in blockable context with a CMA-enabled config is just going to fail. Similarly, it looks like none of the DMA_ATTR_FORCE_CONTIGUOUS cases are prepared to handle this change either (amd_iommu appears technically affected, but is already using dma_alloc_from_contiguous() backwards compared to everyone else, hmm).

Yea, I searched the tree and got the same results.

Quoted 2 lines

I guess the question is whether to add alloc_page()/free_page() fallbacks to those call sites, or stuff them directly into the CMA helpers here.

Probably would be safer/easier to do the later one I feel.

Tested-by on v1 1/1
#4 lore raw
In reply to Nicolin Chen
Quoted 11 lines · Nicolin Chen

This reverts commit d222e42e88168fd67e6d131984b86477af1fc256.

The original change breaks omap dss:
omapdss_dispc 58001000.dispc:
dispc_errata_i734_wa_init: dma_alloc_writecombine failed

Let's revert it first and then find a safer solution instead.

Quoted 7 lines

---
Tony,

Would you please test and verify? Thanks!

 kernel/dma/contiguous.c | 22 +++-------------------
 1 file changed, 3 insertions(+), 19 deletions(-)

This fixes broken (since yesterday) linux-next on Exynos boards (ARMv7), errors like:
dma-pl330: probe of 121a0000.pdma failed with error -12 exynos-ehci 12110000.usb: can't setup: -12 exynos-ehci 12110000.usb: USB bus 1 deregistered exynos-ehci 12110000.usb: Failed to add USB HCD

Best regards,
Krzysztof

maintainer Tested-by on v1 1/1
#5 lore raw
In reply to Nicolin Chen
Quoted 6 lines · Nicolin Chen

This reverts commit d222e42e88168fd67e6d131984b86477af1fc256.

The original change breaks omap dss:
omapdss_dispc 58001000.dispc:
dispc_errata_i734_wa_init: dma_alloc_writecombine failed

Quoted 8 lines

Let's revert it first and then find a safer solution instead.

---
Tony,

Would you please test and verify? Thanks!

This also fixes various memory allocation failures we have seen on 32-bit Tegra as well.

Cheers
Jon

Signature · 1 line
nvpublic