Claire Chang
2021-Apr-22 08:15 UTC
[Nouveau] [PATCH v5 14/16] dma-direct: Allocate memory from restricted DMA pool if available
The restricted DMA pool is preferred if available.
The restricted DMA pools provide a basic level of protection against the
DMA overwriting buffer contents at unexpected times. However, to protect
against general data leakage and system memory corruption, the system
needs to provide a way to lock down the memory access, e.g., MPU.
Signed-off-by: Claire Chang <tientzu at chromium.org>
---
kernel/dma/direct.c | 35 ++++++++++++++++++++++++++---------
1 file changed, 26 insertions(+), 9 deletions(-)
diff --git a/kernel/dma/direct.c b/kernel/dma/direct.c
index 7a27f0510fcc..29523d2a9845 100644
--- a/kernel/dma/direct.c
+++ b/kernel/dma/direct.c
@@ -78,6 +78,10 @@ static bool dma_coherent_ok(struct device *dev, phys_addr_t
phys, size_t size)
static void __dma_direct_free_pages(struct device *dev, struct page *page,
size_t size)
{
+#ifdef CONFIG_DMA_RESTRICTED_POOL
+ if (swiotlb_free(dev, page, size))
+ return;
+#endif
dma_free_contiguous(dev, page, size);
}
@@ -92,7 +96,17 @@ static struct page *__dma_direct_alloc_pages(struct device
*dev, size_t size,
gfp |= dma_direct_optimal_gfp_mask(dev, dev->coherent_dma_mask,
&phys_limit);
- page = dma_alloc_contiguous(dev, size, gfp);
+
+#ifdef CONFIG_DMA_RESTRICTED_POOL
+ page = swiotlb_alloc(dev, size);
+ if (page && !dma_coherent_ok(dev, page_to_phys(page), size)) {
+ __dma_direct_free_pages(dev, page, size);
+ page = NULL;
+ }
+#endif
+
+ if (!page)
+ page = dma_alloc_contiguous(dev, size, gfp);
if (page && !dma_coherent_ok(dev, page_to_phys(page), size)) {
dma_free_contiguous(dev, page, size);
page = NULL;
@@ -148,7 +162,7 @@ void *dma_direct_alloc(struct device *dev, size_t size,
gfp |= __GFP_NOWARN;
if ((attrs & DMA_ATTR_NO_KERNEL_MAPPING) &&
- !force_dma_unencrypted(dev)) {
+ !force_dma_unencrypted(dev) && !is_dev_swiotlb_force(dev)) {
page = __dma_direct_alloc_pages(dev, size, gfp & ~__GFP_ZERO);
if (!page)
return NULL;
@@ -161,8 +175,8 @@ void *dma_direct_alloc(struct device *dev, size_t size,
}
if (!IS_ENABLED(CONFIG_ARCH_HAS_DMA_SET_UNCACHED) &&
- !IS_ENABLED(CONFIG_DMA_DIRECT_REMAP) &&
- !dev_is_dma_coherent(dev))
+ !IS_ENABLED(CONFIG_DMA_DIRECT_REMAP) && !dev_is_dma_coherent(dev)
&&
+ !is_dev_swiotlb_force(dev))
return arch_dma_alloc(dev, size, dma_handle, gfp, attrs);
/*
@@ -172,7 +186,9 @@ void *dma_direct_alloc(struct device *dev, size_t size,
if (IS_ENABLED(CONFIG_DMA_COHERENT_POOL) &&
!gfpflags_allow_blocking(gfp) &&
(force_dma_unencrypted(dev) ||
- (IS_ENABLED(CONFIG_DMA_DIRECT_REMAP) &&
!dev_is_dma_coherent(dev))))
+ (IS_ENABLED(CONFIG_DMA_DIRECT_REMAP) &&
+ !dev_is_dma_coherent(dev))) &&
+ !is_dev_swiotlb_force(dev))
return dma_direct_alloc_from_pool(dev, size, dma_handle, gfp);
/* we always manually zero the memory once we are done */
@@ -253,15 +269,15 @@ void dma_direct_free(struct device *dev, size_t size,
unsigned int page_order = get_order(size);
if ((attrs & DMA_ATTR_NO_KERNEL_MAPPING) &&
- !force_dma_unencrypted(dev)) {
+ !force_dma_unencrypted(dev) && !is_dev_swiotlb_force(dev)) {
/* cpu_addr is a struct page cookie, not a kernel address */
dma_free_contiguous(dev, cpu_addr, size);
return;
}
if (!IS_ENABLED(CONFIG_ARCH_HAS_DMA_SET_UNCACHED) &&
- !IS_ENABLED(CONFIG_DMA_DIRECT_REMAP) &&
- !dev_is_dma_coherent(dev)) {
+ !IS_ENABLED(CONFIG_DMA_DIRECT_REMAP) && !dev_is_dma_coherent(dev)
&&
+ !is_dev_swiotlb_force(dev)) {
arch_dma_free(dev, size, cpu_addr, dma_addr, attrs);
return;
}
@@ -289,7 +305,8 @@ struct page *dma_direct_alloc_pages(struct device *dev,
size_t size,
void *ret;
if (IS_ENABLED(CONFIG_DMA_COHERENT_POOL) &&
- force_dma_unencrypted(dev) && !gfpflags_allow_blocking(gfp))
+ force_dma_unencrypted(dev) && !gfpflags_allow_blocking(gfp)
&&
+ !is_dev_swiotlb_force(dev))
return dma_direct_alloc_from_pool(dev, size, dma_handle, gfp);
page = __dma_direct_alloc_pages(dev, size, gfp);
--
2.31.1.368.gbe11c130af-goog
Robin Murphy
2021-Apr-23 13:46 UTC
[Nouveau] [PATCH v5 14/16] dma-direct: Allocate memory from restricted DMA pool if available
On 2021-04-22 09:15, Claire Chang wrote:> The restricted DMA pool is preferred if available. > > The restricted DMA pools provide a basic level of protection against the > DMA overwriting buffer contents at unexpected times. However, to protect > against general data leakage and system memory corruption, the system > needs to provide a way to lock down the memory access, e.g., MPU. > > Signed-off-by: Claire Chang <tientzu at chromium.org> > --- > kernel/dma/direct.c | 35 ++++++++++++++++++++++++++--------- > 1 file changed, 26 insertions(+), 9 deletions(-) > > diff --git a/kernel/dma/direct.c b/kernel/dma/direct.c > index 7a27f0510fcc..29523d2a9845 100644 > --- a/kernel/dma/direct.c > +++ b/kernel/dma/direct.c > @@ -78,6 +78,10 @@ static bool dma_coherent_ok(struct device *dev, phys_addr_t phys, size_t size) > static void __dma_direct_free_pages(struct device *dev, struct page *page, > size_t size) > { > +#ifdef CONFIG_DMA_RESTRICTED_POOL > + if (swiotlb_free(dev, page, size)) > + return; > +#endif > dma_free_contiguous(dev, page, size); > } > > @@ -92,7 +96,17 @@ static struct page *__dma_direct_alloc_pages(struct device *dev, size_t size, > > gfp |= dma_direct_optimal_gfp_mask(dev, dev->coherent_dma_mask, > &phys_limit); > - page = dma_alloc_contiguous(dev, size, gfp); > + > +#ifdef CONFIG_DMA_RESTRICTED_POOL > + page = swiotlb_alloc(dev, size); > + if (page && !dma_coherent_ok(dev, page_to_phys(page), size)) { > + __dma_direct_free_pages(dev, page, size); > + page = NULL; > + } > +#endif > + > + if (!page) > + page = dma_alloc_contiguous(dev, size, gfp); > if (page && !dma_coherent_ok(dev, page_to_phys(page), size)) { > dma_free_contiguous(dev, page, size); > page = NULL; > @@ -148,7 +162,7 @@ void *dma_direct_alloc(struct device *dev, size_t size, > gfp |= __GFP_NOWARN; > > if ((attrs & DMA_ATTR_NO_KERNEL_MAPPING) && > - !force_dma_unencrypted(dev)) { > + !force_dma_unencrypted(dev) && !is_dev_swiotlb_force(dev)) { > page = __dma_direct_alloc_pages(dev, size, gfp & ~__GFP_ZERO); > if (!page) > return NULL; > @@ -161,8 +175,8 @@ void *dma_direct_alloc(struct device *dev, size_t size, > } > > if (!IS_ENABLED(CONFIG_ARCH_HAS_DMA_SET_UNCACHED) && > - !IS_ENABLED(CONFIG_DMA_DIRECT_REMAP) && > - !dev_is_dma_coherent(dev)) > + !IS_ENABLED(CONFIG_DMA_DIRECT_REMAP) && !dev_is_dma_coherent(dev) && > + !is_dev_swiotlb_force(dev)) > return arch_dma_alloc(dev, size, dma_handle, gfp, attrs); > > /* > @@ -172,7 +186,9 @@ void *dma_direct_alloc(struct device *dev, size_t size, > if (IS_ENABLED(CONFIG_DMA_COHERENT_POOL) && > !gfpflags_allow_blocking(gfp) && > (force_dma_unencrypted(dev) || > - (IS_ENABLED(CONFIG_DMA_DIRECT_REMAP) && !dev_is_dma_coherent(dev)))) > + (IS_ENABLED(CONFIG_DMA_DIRECT_REMAP) && > + !dev_is_dma_coherent(dev))) && > + !is_dev_swiotlb_force(dev)) > return dma_direct_alloc_from_pool(dev, size, dma_handle, gfp); > > /* we always manually zero the memory once we are done */ > @@ -253,15 +269,15 @@ void dma_direct_free(struct device *dev, size_t size, > unsigned int page_order = get_order(size); > > if ((attrs & DMA_ATTR_NO_KERNEL_MAPPING) && > - !force_dma_unencrypted(dev)) { > + !force_dma_unencrypted(dev) && !is_dev_swiotlb_force(dev)) { > /* cpu_addr is a struct page cookie, not a kernel address */ > dma_free_contiguous(dev, cpu_addr, size); > return; > } > > if (!IS_ENABLED(CONFIG_ARCH_HAS_DMA_SET_UNCACHED) && > - !IS_ENABLED(CONFIG_DMA_DIRECT_REMAP) && > - !dev_is_dma_coherent(dev)) { > + !IS_ENABLED(CONFIG_DMA_DIRECT_REMAP) && !dev_is_dma_coherent(dev) && > + !is_dev_swiotlb_force(dev)) { > arch_dma_free(dev, size, cpu_addr, dma_addr, attrs); > return; > } > @@ -289,7 +305,8 @@ struct page *dma_direct_alloc_pages(struct device *dev, size_t size, > void *ret; > > if (IS_ENABLED(CONFIG_DMA_COHERENT_POOL) && > - force_dma_unencrypted(dev) && !gfpflags_allow_blocking(gfp)) > + force_dma_unencrypted(dev) && !gfpflags_allow_blocking(gfp) && > + !is_dev_swiotlb_force(dev)) > return dma_direct_alloc_from_pool(dev, size, dma_handle, gfp);Wait, this seems broken for non-coherent devices - in that case we need to return a non-cacheable address, but we can't simply fall through into the remapping path below in GFP_ATOMIC context. That's why we need the atomic pool concept in the first place :/ Unless I've overlooked something, we're still using the regular cacheable linear map address of the dma_io_tlb_mem buffer, no? Robin.> > page = __dma_direct_alloc_pages(dev, size, gfp); >
Claire Chang
2021-May-03 14:26 UTC
[Nouveau] [PATCH v5 14/16] dma-direct: Allocate memory from restricted DMA pool if available
On Fri, Apr 23, 2021 at 9:46 PM Robin Murphy <robin.murphy at arm.com> wrote:> > On 2021-04-22 09:15, Claire Chang wrote: > > The restricted DMA pool is preferred if available. > > > > The restricted DMA pools provide a basic level of protection against the > > DMA overwriting buffer contents at unexpected times. However, to protect > > against general data leakage and system memory corruption, the system > > needs to provide a way to lock down the memory access, e.g., MPU. > > > > Signed-off-by: Claire Chang <tientzu at chromium.org> > > --- > > kernel/dma/direct.c | 35 ++++++++++++++++++++++++++--------- > > 1 file changed, 26 insertions(+), 9 deletions(-) > > > > diff --git a/kernel/dma/direct.c b/kernel/dma/direct.c > > index 7a27f0510fcc..29523d2a9845 100644 > > --- a/kernel/dma/direct.c > > +++ b/kernel/dma/direct.c > > @@ -78,6 +78,10 @@ static bool dma_coherent_ok(struct device *dev, phys_addr_t phys, size_t size) > > static void __dma_direct_free_pages(struct device *dev, struct page *page, > > size_t size) > > { > > +#ifdef CONFIG_DMA_RESTRICTED_POOL > > + if (swiotlb_free(dev, page, size)) > > + return; > > +#endif > > dma_free_contiguous(dev, page, size); > > } > > > > @@ -92,7 +96,17 @@ static struct page *__dma_direct_alloc_pages(struct device *dev, size_t size, > > > > gfp |= dma_direct_optimal_gfp_mask(dev, dev->coherent_dma_mask, > > &phys_limit); > > - page = dma_alloc_contiguous(dev, size, gfp); > > + > > +#ifdef CONFIG_DMA_RESTRICTED_POOL > > + page = swiotlb_alloc(dev, size); > > + if (page && !dma_coherent_ok(dev, page_to_phys(page), size)) { > > + __dma_direct_free_pages(dev, page, size); > > + page = NULL; > > + } > > +#endif > > + > > + if (!page) > > + page = dma_alloc_contiguous(dev, size, gfp); > > if (page && !dma_coherent_ok(dev, page_to_phys(page), size)) { > > dma_free_contiguous(dev, page, size); > > page = NULL; > > @@ -148,7 +162,7 @@ void *dma_direct_alloc(struct device *dev, size_t size, > > gfp |= __GFP_NOWARN; > > > > if ((attrs & DMA_ATTR_NO_KERNEL_MAPPING) && > > - !force_dma_unencrypted(dev)) { > > + !force_dma_unencrypted(dev) && !is_dev_swiotlb_force(dev)) { > > page = __dma_direct_alloc_pages(dev, size, gfp & ~__GFP_ZERO); > > if (!page) > > return NULL; > > @@ -161,8 +175,8 @@ void *dma_direct_alloc(struct device *dev, size_t size, > > } > > > > if (!IS_ENABLED(CONFIG_ARCH_HAS_DMA_SET_UNCACHED) && > > - !IS_ENABLED(CONFIG_DMA_DIRECT_REMAP) && > > - !dev_is_dma_coherent(dev)) > > + !IS_ENABLED(CONFIG_DMA_DIRECT_REMAP) && !dev_is_dma_coherent(dev) && > > + !is_dev_swiotlb_force(dev)) > > return arch_dma_alloc(dev, size, dma_handle, gfp, attrs); > > > > /* > > @@ -172,7 +186,9 @@ void *dma_direct_alloc(struct device *dev, size_t size, > > if (IS_ENABLED(CONFIG_DMA_COHERENT_POOL) && > > !gfpflags_allow_blocking(gfp) && > > (force_dma_unencrypted(dev) || > > - (IS_ENABLED(CONFIG_DMA_DIRECT_REMAP) && !dev_is_dma_coherent(dev)))) > > + (IS_ENABLED(CONFIG_DMA_DIRECT_REMAP) && > > + !dev_is_dma_coherent(dev))) && > > + !is_dev_swiotlb_force(dev)) > > return dma_direct_alloc_from_pool(dev, size, dma_handle, gfp); > > > > /* we always manually zero the memory once we are done */ > > @@ -253,15 +269,15 @@ void dma_direct_free(struct device *dev, size_t size, > > unsigned int page_order = get_order(size); > > > > if ((attrs & DMA_ATTR_NO_KERNEL_MAPPING) && > > - !force_dma_unencrypted(dev)) { > > + !force_dma_unencrypted(dev) && !is_dev_swiotlb_force(dev)) { > > /* cpu_addr is a struct page cookie, not a kernel address */ > > dma_free_contiguous(dev, cpu_addr, size); > > return; > > } > > > > if (!IS_ENABLED(CONFIG_ARCH_HAS_DMA_SET_UNCACHED) && > > - !IS_ENABLED(CONFIG_DMA_DIRECT_REMAP) && > > - !dev_is_dma_coherent(dev)) { > > + !IS_ENABLED(CONFIG_DMA_DIRECT_REMAP) && !dev_is_dma_coherent(dev) && > > + !is_dev_swiotlb_force(dev)) { > > arch_dma_free(dev, size, cpu_addr, dma_addr, attrs); > > return; > > } > > @@ -289,7 +305,8 @@ struct page *dma_direct_alloc_pages(struct device *dev, size_t size, > > void *ret; > > > > if (IS_ENABLED(CONFIG_DMA_COHERENT_POOL) && > > - force_dma_unencrypted(dev) && !gfpflags_allow_blocking(gfp)) > > + force_dma_unencrypted(dev) && !gfpflags_allow_blocking(gfp) && > > + !is_dev_swiotlb_force(dev)) > > return dma_direct_alloc_from_pool(dev, size, dma_handle, gfp); > > Wait, this seems broken for non-coherent devices - in that case we need > to return a non-cacheable address, but we can't simply fall through into > the remapping path below in GFP_ATOMIC context. That's why we need the > atomic pool concept in the first place :/Sorry for the late reply. I'm not very familiar with this. I wonder if the memory returned here must be coherent. If yes, could we say for this case, one must set up another device coherent pool (shared-dma-pool) and go with dma_alloc_from_dev_coherent()[1]? [1] https://elixir.bootlin.com/linux/v5.12/source/kernel/dma/mapping.c#L435> > Unless I've overlooked something, we're still using the regular > cacheable linear map address of the dma_io_tlb_mem buffer, no? > > Robin. > > > > > page = __dma_direct_alloc_pages(dev, size, gfp); > >