| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Tue, 20 Aug 2019 22:24:29 +0200
From: Tobias Klausmann <tobias.johannes.klausmann@....thm.de>
To: Christoph Hellwig <hch@....de>, Hillf Danton <hdanton@...a.com>
Cc: Nicolin Chen <nicoleotsuka@...il.com>, kvalo@...eaurora.org,
davem@...emloft.net, ath10k@...ts.infradead.org,
linux-wireless@...r.kernel.org, netdev@...r.kernel.org,
linux-kernel@...r.kernel.org, m.szyprowski@...sung.com,
robin.murphy@....com, iommu@...ts.linux-foundation.org,
tobias.klausmann@...enet.de
Subject: Re: regression in ath10k dma allocation
On 20.08.19 09:12, Christoph Hellwig wrote:
> On Tue, Aug 20, 2019 at 02:58:33PM +0800, Hillf Danton wrote:
>> On Tue, 20 Aug 2019 05:05:14 +0200 Christoph Hellwig wrote:
>>> Tobias, plase try this patch:
>>>
> New version below:
>
> ---
> From b8a805e93be5a5662323b8ac61fe686df839c4ac Mon Sep 17 00:00:00 2001
> From: Christoph Hellwig <hch@....de>
> Date: Tue, 20 Aug 2019 11:45:49 +0900
> Subject: dma-direct: fix zone selection after an unaddressable CMA allocation
>
> The new dma_alloc_contiguous hides if we allocate CMA or regular
> pages, and thus fails to retry a ZONE_NORMAL allocation if the CMA
> allocation succeeds but isn't addressable. That means we either fail
> outright or dip into a small zone that might not succeed either.
>
> Thanks to Hillf Danton for debugging this issue.
>
> Fixes: b1d2dc009dec ("dma-contiguous: add dma_{alloc,free}_contiguous() helpers")
> Reported-by: Tobias Klausmann <tobias.johannes.klausmann@....thm.de>
> Signed-off-by: Christoph Hellwig <hch@....de>
> ---
> drivers/iommu/dma-iommu.c | 3 +++
> include/linux/dma-contiguous.h | 5 +----
> kernel/dma/contiguous.c | 9 +++------
> kernel/dma/direct.c | 10 +++++++++-
> 4 files changed, 16 insertions(+), 11 deletions(-)
>
> diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
> index d991d40f797f..f68a62c3c32b 100644
> --- a/drivers/iommu/dma-iommu.c
> +++ b/drivers/iommu/dma-iommu.c
> @@ -965,10 +965,13 @@ static void *iommu_dma_alloc_pages(struct device *dev, size_t size,
> {
> bool coherent = dev_is_dma_coherent(dev);
> size_t alloc_size = PAGE_ALIGN(size);
> + int node = dev_to_node(dev);
> struct page *page = NULL;
> void *cpu_addr;
>
> page = dma_alloc_contiguous(dev, alloc_size, gfp);
> + if (!page)
> + page = alloc_pages_node(node, gfp, get_order(alloc_size));
> if (!page)
> return NULL;
>
> diff --git a/include/linux/dma-contiguous.h b/include/linux/dma-contiguous.h
> index c05d4e661489..03f8e98e3bcc 100644
> --- a/include/linux/dma-contiguous.h
> +++ b/include/linux/dma-contiguous.h
> @@ -160,10 +160,7 @@ bool dma_release_from_contiguous(struct device *dev, struct page *pages,
> static inline struct page *dma_alloc_contiguous(struct device *dev, size_t size,
> gfp_t gfp)
> {
> - int node = dev ? dev_to_node(dev) : NUMA_NO_NODE;
> - size_t align = get_order(PAGE_ALIGN(size));
> -
> - return alloc_pages_node(node, gfp, align);
> + return NULL;
> }
>
> static inline void dma_free_contiguous(struct device *dev, struct page *page,
> diff --git a/kernel/dma/contiguous.c b/kernel/dma/contiguous.c
> index 2bd410f934b3..e6b450fdbeb6 100644
> --- a/kernel/dma/contiguous.c
> +++ b/kernel/dma/contiguous.c
> @@ -230,9 +230,7 @@ bool dma_release_from_contiguous(struct device *dev, struct page *pages,
> */
> struct page *dma_alloc_contiguous(struct device *dev, size_t size, gfp_t gfp)
> {
> - int node = dev ? dev_to_node(dev) : NUMA_NO_NODE;
> - size_t count = PAGE_ALIGN(size) >> PAGE_SHIFT;
> - size_t align = get_order(PAGE_ALIGN(size));
> + size_t count = size >> PAGE_SHIFT;
> struct page *page = NULL;
> struct cma *cma = NULL;
>
> @@ -243,14 +241,12 @@ struct page *dma_alloc_contiguous(struct device *dev, size_t size, gfp_t gfp)
>
> /* CMA can be used only in the context which permits sleeping */
> if (cma && gfpflags_allow_blocking(gfp)) {
> + size_t align = get_order(size);
> size_t cma_align = min_t(size_t, align, CONFIG_CMA_ALIGNMENT);
>
> page = cma_alloc(cma, count, cma_align, gfp & __GFP_NOWARN);
> }
>
> - /* Fallback allocation of normal pages */
> - if (!page)
> - page = alloc_pages_node(node, gfp, align);
> return page;
> }
>
> @@ -258,6 +254,7 @@ struct page *dma_alloc_contiguous(struct device *dev, size_t size, gfp_t gfp)
> * dma_free_contiguous() - release allocated pages
> * @dev: Pointer to device for which the pages were allocated.
> * @page: Pointer to the allocated pages.
> + int node = dev ? dev_to_node(dev) : NUMA_NO_NODE;
> * @size: Size of allocated pages.
> *
> * This function releases memory allocated by dma_alloc_contiguous(). As the
> diff --git a/kernel/dma/direct.c b/kernel/dma/direct.c
> index 795c9b095d75..706113c6bebc 100644
> --- a/kernel/dma/direct.c
> +++ b/kernel/dma/direct.c
> @@ -85,6 +85,8 @@ static bool dma_coherent_ok(struct device *dev, phys_addr_t phys, size_t size)
> struct page *__dma_direct_alloc_pages(struct device *dev, size_t size,
> dma_addr_t *dma_handle, gfp_t gfp, unsigned long attrs)
> {
> + size_t alloc_size = PAGE_ALIGN(size);
> + int node = dev_to_node(dev);
> struct page *page = NULL;
> u64 phys_mask;
>
> @@ -95,8 +97,14 @@ struct page *__dma_direct_alloc_pages(struct device *dev, size_t size,
> gfp &= ~__GFP_ZERO;
> gfp |= __dma_direct_optimal_gfp_mask(dev, dev->coherent_dma_mask,
> &phys_mask);
> + page = dma_alloc_contiguous(dev, alloc_size, gfp);
> + if (page && !dma_coherent_ok(dev, page_to_phys(page), size)) {
> + dma_free_contiguous(dev, page, alloc_size);
> + page = NULL;
> + }
> again:
> - page = dma_alloc_contiguous(dev, size, gfp);
> + if (!page)
> + page = alloc_pages_node(node, gfp, get_order(alloc_size));
> if (page && !dma_coherent_ok(dev, page_to_phys(page), size)) {
> dma_free_contiguous(dev, page, size);
> page = NULL;
I can confirm this resolves the regression!
Tested-by: Tobias Klausmann <tobias.johannes.klausmann@....thm.de>
Thanks for the work of all involved,
Tobias
Powered by blists - more mailing lists