* [PATCH] dma-mapping: don't clear GFP_ZERO in dma_alloc_attrs
@ 2018-03-28 13:35 ` Christoph Hellwig
0 siblings, 0 replies; 11+ messages in thread
From: Christoph Hellwig @ 2018-03-28 13:35 UTC (permalink / raw)
To: x86-DgEjT+Ai2ygdnm+yROfE0A
Cc: Evgeniy Didin, iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA,
Sebastian Ott, linux-kernel-u79uwXL29TY76Z2rM5mHXA
Revert the clearing of __GFP_ZERO in dma_alloc_attrs and move it to
dma_direct_alloc for now. While most common architectures always zero dma
cohereny allocations (and x86 did so since day one) this is not documented
and at least arc and s390 do not zero without the explicit __GFP_ZERO
argument.
Fixes: 57bf5a8963f8 ("dma-mapping: clear harmful GFP_* flags in common code")
Reported-by: Evgeniy Didin <Evgeniy.Didin-HKixBCOQz3hWk0Htik3J/w@public.gmane.org>
Reported-by: Sebastian Ott <sebott-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
---
include/linux/dma-mapping.h | 8 ++------
lib/dma-direct.c | 3 +++
2 files changed, 5 insertions(+), 6 deletions(-)
diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h
index eb9eab4ecd6d..12fedcba9a9a 100644
--- a/include/linux/dma-mapping.h
+++ b/include/linux/dma-mapping.h
@@ -518,12 +518,8 @@ static inline void *dma_alloc_attrs(struct device *dev, size_t size,
if (dma_alloc_from_dev_coherent(dev, size, dma_handle, &cpu_addr))
return cpu_addr;
- /*
- * Let the implementation decide on the zone to allocate from, and
- * decide on the way of zeroing the memory given that the memory
- * returned should always be zeroed.
- */
- flag &= ~(__GFP_DMA | __GFP_DMA32 | __GFP_HIGHMEM | __GFP_ZERO);
+ /* let the implementation decide on the zone to allocate from: */
+ flag &= ~(__GFP_DMA | __GFP_DMA32 | __GFP_HIGHMEM);
if (!arch_dma_alloc_attrs(&dev, &flag))
return NULL;
diff --git a/lib/dma-direct.c b/lib/dma-direct.c
index 1277d293d4da..c0bba30fef0a 100644
--- a/lib/dma-direct.c
+++ b/lib/dma-direct.c
@@ -59,6 +59,9 @@ void *dma_direct_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle,
struct page *page = NULL;
void *ret;
+ /* we always manually zero the memory once we are done: */
+ gfp &= ~__GFP_ZERO;
+
/* GFP_DMA32 and GFP_DMA are no ops without the corresponding zones: */
if (dev->coherent_dma_mask <= DMA_BIT_MASK(ARCH_ZONE_DMA_BITS))
gfp |= GFP_DMA;
--
2.14.2
^ permalink raw reply related [flat|nested] 11+ messages in thread* Re: [PATCH] dma-mapping: don't clear GFP_ZERO in dma_alloc_attrs
@ 2018-03-28 15:12 ` Evgeniy Didin
0 siblings, 0 replies; 11+ messages in thread
From: Evgeniy Didin @ 2018-03-28 15:12 UTC (permalink / raw)
To: hch, x86
Cc: linux-kernel, sebott, linux-arch, iommu, Evgeniy.Didin, linux-snps-arc
Adding linux-snps and linux-arch mailing lists.
> Revert the clearing of __GFP_ZERO in dma_alloc_attrs and move it to
> dma_direct_alloc for now. While most common architectures always zero dma
> cohereny allocations (and x86 did so since day one) this is not documented
> and at least arc and s390 do not zero without the explicit __GFP_ZERO
> argument.
This patch fixed Ethernet issues on ARC HSDK.
https://www.spinics.net/lists/kernel/msg2762054.html
Tested-by: Evgeniy Didin <Evgeniy.Didin@synopsys.com>
> Fixes: 57bf5a8963f8 ("dma-mapping: clear harmful GFP_* flags in common code")
> Reported-by: Evgeniy Didin <Evgeniy.Didin@synopsys.com>
> Reported-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
> include/linux/dma-mapping.h | 8 ++------
> lib/dma-direct.c | 3 +++
> 2 files changed, 5 insertions(+), 6 deletions(-)
>
> diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h
> index eb9eab4ecd6d..12fedcba9a9a 100644
> --- a/include/linux/dma-mapping.h
> +++ b/include/linux/dma-mapping.h
> @@ -518,12 +518,8 @@ static inline void *dma_alloc_attrs(struct device *dev, size_t size,
> if (dma_alloc_from_dev_coherent(dev, size, dma_handle, &cpu_addr))
> return cpu_addr;
>
> - /*
> - * Let the implementation decide on the zone to allocate from, and
> - * decide on the way of zeroing the memory given that the memory
> - * returned should always be zeroed.
> - */
> - flag &= ~(__GFP_DMA | __GFP_DMA32 | __GFP_HIGHMEM | __GFP_ZERO);
> + /* let the implementation decide on the zone to allocate from: */
> + flag &= ~(__GFP_DMA | __GFP_DMA32 | __GFP_HIGHMEM);
>
> if (!arch_dma_alloc_attrs(&dev, &flag))
> return NULL;
> diff --git a/lib/dma-direct.c b/lib/dma-direct.c
> index 1277d293d4da..c0bba30fef0a 100644
> --- a/lib/dma-direct.c
> +++ b/lib/dma-direct.c
> @@ -59,6 +59,9 @@ void *dma_direct_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle,
> struct page *page = NULL;
> void *ret;
>
> + /* we always manually zero the memory once we are done: */
> + gfp &= ~__GFP_ZERO;
> +
> /* GFP_DMA32 and GFP_DMA are no ops without the corresponding zones: */
> if (dev->coherent_dma_mask <= DMA_BIT_MASK(ARCH_ZONE_DMA_BITS))
> gfp |= GFP_DMA;
^ permalink raw reply [flat|nested] 11+ messages in thread* [PATCH] dma-mapping: don't clear GFP_ZERO in dma_alloc_attrs
@ 2018-03-28 15:12 ` Evgeniy Didin
0 siblings, 0 replies; 11+ messages in thread
From: Evgeniy Didin @ 2018-03-28 15:12 UTC (permalink / raw)
To: linux-snps-arc
Adding linux-snps and linux-arch mailing lists.
> Revert the clearing of __GFP_ZERO in dma_alloc_attrs and move it to
> dma_direct_alloc for now.??While most common architectures always zero dma
> cohereny allocations (and x86 did so since day one) this is not documented
> and at least arc and s390 do not zero without the explicit __GFP_ZERO
> argument.
This patch fixed Ethernet issues on ARC HSDK.
https://www.spinics.net/lists/kernel/msg2762054.html
Tested-by: Evgeniy Didin <Evgeniy.Didin at synopsys.com>
> Fixes: 57bf5a8963f8 ("dma-mapping: clear harmful GFP_* flags in common code")
> Reported-by: Evgeniy Didin <Evgeniy.Didin at synopsys.com>
> Reported-by: Sebastian Ott <sebott at linux.vnet.ibm.com>
> Signed-off-by: Christoph Hellwig <hch at lst.de>
> ---
> ?include/linux/dma-mapping.h | 8 ++------
> ?lib/dma-direct.c????????????| 3 +++
> ?2 files changed, 5 insertions(+), 6 deletions(-)
>
> diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h
> index eb9eab4ecd6d..12fedcba9a9a 100644
> --- a/include/linux/dma-mapping.h
> +++ b/include/linux/dma-mapping.h
> @@ -518,12 +518,8 @@ static inline void *dma_alloc_attrs(struct device *dev, size_t size,
> ? if (dma_alloc_from_dev_coherent(dev, size, dma_handle, &cpu_addr))
> ? return cpu_addr;
> ?
> - /*
> - ?* Let the implementation decide on the zone to allocate from, and
> - ?* decide on the way of zeroing the memory given that the memory
> - ?* returned should always be zeroed.
> - ?*/
> - flag &= ~(__GFP_DMA | __GFP_DMA32 | __GFP_HIGHMEM | __GFP_ZERO);
> + /* let the implementation decide on the zone to allocate from: */
> + flag &= ~(__GFP_DMA | __GFP_DMA32 | __GFP_HIGHMEM);
>
> ? if (!arch_dma_alloc_attrs(&dev, &flag))
> ? return NULL;
> diff --git a/lib/dma-direct.c b/lib/dma-direct.c
> index 1277d293d4da..c0bba30fef0a 100644
> --- a/lib/dma-direct.c
> +++ b/lib/dma-direct.c
> @@ -59,6 +59,9 @@ void *dma_direct_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle,
> ? struct page *page = NULL;
> ? void *ret;
> ?
> + /* we always manually zero the memory once we are done: */
> + gfp &= ~__GFP_ZERO;
> +
> ? /* GFP_DMA32 and GFP_DMA are no ops without the corresponding zones: */
> ? if (dev->coherent_dma_mask <= DMA_BIT_MASK(ARCH_ZONE_DMA_BITS))
> ? gfp |= GFP_DMA;
^ permalink raw reply [flat|nested] 11+ messages in thread* Re: [PATCH] dma-mapping: don't clear GFP_ZERO in dma_alloc_attrs
@ 2018-03-28 15:12 ` Evgeniy Didin
0 siblings, 0 replies; 11+ messages in thread
From: Evgeniy Didin @ 2018-03-28 15:12 UTC (permalink / raw)
To: hch-jcswGhMUV9g, x86-DgEjT+Ai2ygdnm+yROfE0A
Cc: linux-arch-u79uwXL29TY76Z2rM5mHXA,
sebott-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8,
linux-kernel-u79uwXL29TY76Z2rM5mHXA,
Evgeniy.Didin-HKixBCOQz3hWk0Htik3J/w,
iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA,
linux-snps-arc-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r
Adding linux-snps and linux-arch mailing lists.
> Revert the clearing of __GFP_ZERO in dma_alloc_attrs and move it to
> dma_direct_alloc for now. While most common architectures always zero dma
> cohereny allocations (and x86 did so since day one) this is not documented
> and at least arc and s390 do not zero without the explicit __GFP_ZERO
> argument.
This patch fixed Ethernet issues on ARC HSDK.
https://www.spinics.net/lists/kernel/msg2762054.html
Tested-by: Evgeniy Didin <Evgeniy.Didin@synopsys.com>
> Fixes: 57bf5a8963f8 ("dma-mapping: clear harmful GFP_* flags in common code")
> Reported-by: Evgeniy Didin <Evgeniy.Didin@synopsys.com>
> Reported-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
> include/linux/dma-mapping.h | 8 ++------
> lib/dma-direct.c | 3 +++
> 2 files changed, 5 insertions(+), 6 deletions(-)
>
> diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h
> index eb9eab4ecd6d..12fedcba9a9a 100644
> --- a/include/linux/dma-mapping.h
> +++ b/include/linux/dma-mapping.h
> @@ -518,12 +518,8 @@ static inline void *dma_alloc_attrs(struct device *dev, size_t size,
> if (dma_alloc_from_dev_coherent(dev, size, dma_handle, &cpu_addr))
> return cpu_addr;
>
> - /*
> - * Let the implementation decide on the zone to allocate from, and
> - * decide on the way of zeroing the memory given that the memory
> - * returned should always be zeroed.
> - */
> - flag &= ~(__GFP_DMA | __GFP_DMA32 | __GFP_HIGHMEM | __GFP_ZERO);
> + /* let the implementation decide on the zone to allocate from: */
> + flag &= ~(__GFP_DMA | __GFP_DMA32 | __GFP_HIGHMEM);
>
> if (!arch_dma_alloc_attrs(&dev, &flag))
> return NULL;
> diff --git a/lib/dma-direct.c b/lib/dma-direct.c
> index 1277d293d4da..c0bba30fef0a 100644
> --- a/lib/dma-direct.c
> +++ b/lib/dma-direct.c
> @@ -59,6 +59,9 @@ void *dma_direct_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle,
> struct page *page = NULL;
> void *ret;
>
> + /* we always manually zero the memory once we are done: */
> + gfp &= ~__GFP_ZERO;
> +
> /* GFP_DMA32 and GFP_DMA are no ops without the corresponding zones: */
> if (dev->coherent_dma_mask <= DMA_BIT_MASK(ARCH_ZONE_DMA_BITS))
> gfp |= GFP_DMA;
_______________________________________________
iommu mailing list
iommu@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/iommu
^ permalink raw reply [flat|nested] 11+ messages in thread* dma-mapping: clear harmful GFP_* flags in common code
@ 2018-04-09 10:52 ` Alexey Brodkin
0 siblings, 0 replies; 11+ messages in thread
From: Alexey Brodkin @ 2018-04-09 10:52 UTC (permalink / raw)
To: stable; +Cc: hch, linux-arch, sebott, x86, linux-kernel, iommu, linux-snps-arc
Hello,
May we have
e89f5b370153 ("dma-mapping: Don't clear GFP_ZERO in dma_alloc_attrs")
back-ported to 4.16 kernel as it fixes:
57bf5a8 ("dma-mapping: clear harmful GFP_* flags in common code").
For more info about introduced problem see this thread:
http://lists.infradead.org/pipermail/linux-snps-arc/2018-March/003652.html
-Alexey
^ permalink raw reply [flat|nested] 11+ messages in thread* dma-mapping: clear harmful GFP_* flags in common code
@ 2018-04-09 10:52 ` Alexey Brodkin
0 siblings, 0 replies; 11+ messages in thread
From: Alexey Brodkin @ 2018-04-09 10:52 UTC (permalink / raw)
To: linux-snps-arc
Hello,
May we have
e89f5b370153 ("dma-mapping: Don't clear GFP_ZERO in dma_alloc_attrs")
back-ported to 4.16 kernel as it fixes:
57bf5a8 ("dma-mapping: clear harmful GFP_* flags in common code").
For more info about introduced problem see this thread:
http://lists.infradead.org/pipermail/linux-snps-arc/2018-March/003652.html
-Alexey
^ permalink raw reply [flat|nested] 11+ messages in thread* dma-mapping: clear harmful GFP_* flags in common code
@ 2018-04-09 10:52 ` Alexey Brodkin
0 siblings, 0 replies; 11+ messages in thread
From: Alexey Brodkin @ 2018-04-09 10:52 UTC (permalink / raw)
To: stable-u79uwXL29TY76Z2rM5mHXA
Cc: linux-arch-u79uwXL29TY76Z2rM5mHXA, x86-DgEjT+Ai2ygdnm+yROfE0A,
sebott-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8,
linux-kernel-u79uwXL29TY76Z2rM5mHXA,
iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA,
linux-snps-arc-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r, hch-jcswGhMUV9g
Hello,
May we have
e89f5b370153 ("dma-mapping: Don't clear GFP_ZERO in dma_alloc_attrs")
back-ported to 4.16 kernel as it fixes:
57bf5a8 ("dma-mapping: clear harmful GFP_* flags in common code").
For more info about introduced problem see this thread:
http://lists.infradead.org/pipermail/linux-snps-arc/2018-March/003652.html
-Alexey
^ permalink raw reply [flat|nested] 11+ messages in thread
* [tip:x86/dma] dma-mapping: Don't clear GFP_ZERO in dma_alloc_attrs
2018-03-28 13:35 ` Christoph Hellwig
(?)
(?)
@ 2018-03-28 15:40 ` tip-bot for Christoph Hellwig
-1 siblings, 0 replies; 11+ messages in thread
From: tip-bot for Christoph Hellwig @ 2018-03-28 15:40 UTC (permalink / raw)
To: linux-tip-commits
Cc: sebott, tglx, hch, mingo, Evgeniy.Didin, linux-kernel, hpa
Commit-ID: e89f5b37015309a8bdf0b21d08007580b92f92a4
Gitweb: https://git.kernel.org/tip/e89f5b37015309a8bdf0b21d08007580b92f92a4
Author: Christoph Hellwig <hch@lst.de>
AuthorDate: Wed, 28 Mar 2018 15:35:35 +0200
Committer: Thomas Gleixner <tglx@linutronix.de>
CommitDate: Wed, 28 Mar 2018 17:34:23 +0200
dma-mapping: Don't clear GFP_ZERO in dma_alloc_attrs
Revert the clearing of __GFP_ZERO in dma_alloc_attrs and move it to
dma_direct_alloc for now. While most common architectures always zero dma
cohereny allocations (and x86 did so since day one) this is not documented
and at least arc and s390 do not zero without the explicit __GFP_ZERO
argument.
Fixes: 57bf5a8963f8 ("dma-mapping: clear harmful GFP_* flags in common code")
Reported-by: Evgeniy Didin <Evgeniy.Didin@synopsys.com>
Reported-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Evgeniy Didin <Evgeniy.Didin@synopsys.com>
Cc: iommu@lists.linux-foundation.org
Link: https://lkml.kernel.org/r/20180328133535.17302-2-hch@lst.de
---
include/linux/dma-mapping.h | 8 ++------
lib/dma-direct.c | 3 +++
2 files changed, 5 insertions(+), 6 deletions(-)
diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h
index eb9eab4ecd6d..12fedcba9a9a 100644
--- a/include/linux/dma-mapping.h
+++ b/include/linux/dma-mapping.h
@@ -518,12 +518,8 @@ static inline void *dma_alloc_attrs(struct device *dev, size_t size,
if (dma_alloc_from_dev_coherent(dev, size, dma_handle, &cpu_addr))
return cpu_addr;
- /*
- * Let the implementation decide on the zone to allocate from, and
- * decide on the way of zeroing the memory given that the memory
- * returned should always be zeroed.
- */
- flag &= ~(__GFP_DMA | __GFP_DMA32 | __GFP_HIGHMEM | __GFP_ZERO);
+ /* let the implementation decide on the zone to allocate from: */
+ flag &= ~(__GFP_DMA | __GFP_DMA32 | __GFP_HIGHMEM);
if (!arch_dma_alloc_attrs(&dev, &flag))
return NULL;
diff --git a/lib/dma-direct.c b/lib/dma-direct.c
index 1277d293d4da..c0bba30fef0a 100644
--- a/lib/dma-direct.c
+++ b/lib/dma-direct.c
@@ -59,6 +59,9 @@ void *dma_direct_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle,
struct page *page = NULL;
void *ret;
+ /* we always manually zero the memory once we are done: */
+ gfp &= ~__GFP_ZERO;
+
/* GFP_DMA32 and GFP_DMA are no ops without the corresponding zones: */
if (dev->coherent_dma_mask <= DMA_BIT_MASK(ARCH_ZONE_DMA_BITS))
gfp |= GFP_DMA;
^ permalink raw reply related [flat|nested] 11+ messages in thread