* Re: [PATCH] media: intel/ipu6: Fix dma mask for non-secure mode
2025-04-09 9:58 [PATCH] media: intel/ipu6: Fix dma mask for non-secure mode Stanislaw Gruszka
@ 2025-04-09 10:20 ` Stanislaw Gruszka
2025-04-09 12:20 ` kernel test robot
` (2 subsequent siblings)
3 siblings, 0 replies; 6+ messages in thread
From: Stanislaw Gruszka @ 2025-04-09 10:20 UTC (permalink / raw)
To: linux-media; +Cc: Sakari Ailus, Bingbu Cao
On Wed, Apr 09, 2025 at 11:58:25AM +0200, Stanislaw Gruszka wrote:
> We use dma_get_mask() of auxdev device for calculate iova pfn limit.
> This is always 32 bit mask as we do not initialize the mask (and we can
> not do so, since dev->dev_mask is NULL anyways for auxdev).
>
> Since we need 31 bit mask for non-secure mode create wrapper of
> alloc_iova() which use mmu_info->aperture_end. This give us always
> the correct mask.
>
> Fixes: daabc5c64703 ("media: ipu6: not override the dma_ops of device in driver")
> Cc: stable@vger.kernel.org
> Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com>
<snip>
> void __iomem *base, int mmid,
> const struct ipu6_hw_variants *hw);
> @@ -70,4 +76,11 @@ void ipu6_mmu_unmap(struct ipu6_mmu_info *mmu_info, unsigned long iova,
> size_t size);
> phys_addr_t ipu6_mmu_iova_to_phys(struct ipu6_mmu_info *mmu_info,
> dma_addr_t iova);
> +
> +static inline struct iova *ipu_alloc_iova(struct ipu6_mmu *mmu,
> + unsigned long n_pages)
Before posing one of my vim macro accidentally mangled this patch.
Will send v2.
Regards
Stanislaw
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] media: intel/ipu6: Fix dma mask for non-secure mode
2025-04-09 9:58 [PATCH] media: intel/ipu6: Fix dma mask for non-secure mode Stanislaw Gruszka
2025-04-09 10:20 ` Stanislaw Gruszka
@ 2025-04-09 12:20 ` kernel test robot
2025-04-09 13:14 ` kernel test robot
2025-04-10 6:53 ` Bingbu Cao
3 siblings, 0 replies; 6+ messages in thread
From: kernel test robot @ 2025-04-09 12:20 UTC (permalink / raw)
To: Stanislaw Gruszka, linux-media; +Cc: oe-kbuild-all, Sakari Ailus, Bingbu Cao
Hi Stanislaw,
kernel test robot noticed the following build errors:
[auto build test ERROR on linuxtv-media-pending/master]
[also build test ERROR on linus/master v6.15-rc1 next-20250409]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]
url: https://github.com/intel-lab-lkp/linux/commits/Stanislaw-Gruszka/media-intel-ipu6-Fix-dma-mask-for-non-secure-mode/20250409-180044
base: https://git.linuxtv.org/media-ci/media-pending.git master
patch link: https://lore.kernel.org/r/20250409095825.1014521-1-stanislaw.gruszka%40linux.intel.com
patch subject: [PATCH] media: intel/ipu6: Fix dma mask for non-secure mode
config: i386-buildonly-randconfig-001-20250409 (https://download.01.org/0day-ci/archive/20250409/202504092031.rZzXFFce-lkp@intel.com/config)
compiler: gcc-12 (Debian 12.2.0-14) 12.2.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20250409/202504092031.rZzXFFce-lkp@intel.com/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202504092031.rZzXFFce-lkp@intel.com/
All error/warnings (new ones prefixed by >>):
drivers/media/pci/intel/ipu6/ipu6-dma.c: In function 'ipu6_dma_alloc':
>> drivers/media/pci/intel/ipu6/ipu6-dma.c:174:16: error: implicit declaration of function 'ipu6_alloc_iova'; did you mean 'ipu_alloc_iova'? [-Werror=implicit-function-declaration]
174 | iova = ipu6_alloc_iova(mmu, count);
| ^~~~~~~~~~~~~~~
| ipu_alloc_iova
>> drivers/media/pci/intel/ipu6/ipu6-dma.c:174:14: warning: assignment to 'struct iova *' from 'int' makes pointer from integer without a cast [-Wint-conversion]
174 | iova = ipu6_alloc_iova(mmu, count);
| ^
drivers/media/pci/intel/ipu6/ipu6-dma.c: In function 'ipu6_dma_map_sg':
drivers/media/pci/intel/ipu6/ipu6-dma.c:399:14: warning: assignment to 'struct iova *' from 'int' makes pointer from integer without a cast [-Wint-conversion]
399 | iova = ipu6_alloc_iova(mmu, npages);
| ^
cc1: some warnings being treated as errors
--
drivers/media/pci/intel/ipu6/ipu6-mmu.c: In function 'allocate_trash_buffer':
>> drivers/media/pci/intel/ipu6/ipu6-mmu.c:424:16: error: implicit declaration of function 'ipu6_alloc_iova'; did you mean 'ipu_alloc_iova'? [-Werror=implicit-function-declaration]
424 | iova = ipu6_alloc_iova(mmu, n_pages);
| ^~~~~~~~~~~~~~~
| ipu_alloc_iova
>> drivers/media/pci/intel/ipu6/ipu6-mmu.c:424:14: warning: assignment to 'struct iova *' from 'int' makes pointer from integer without a cast [-Wint-conversion]
424 | iova = ipu6_alloc_iova(mmu, n_pages);
| ^
cc1: some warnings being treated as errors
vim +174 drivers/media/pci/intel/ipu6/ipu6-dma.c
151
152 void *ipu6_dma_alloc(struct ipu6_bus_device *sys, size_t size,
153 dma_addr_t *dma_handle, gfp_t gfp,
154 unsigned long attrs)
155 {
156 struct device *dev = &sys->auxdev.dev;
157 struct pci_dev *pdev = sys->isp->pdev;
158 dma_addr_t pci_dma_addr, ipu6_iova;
159 struct ipu6_mmu *mmu = sys->mmu;
160 struct vm_info *info;
161 unsigned long count;
162 struct page **pages;
163 struct iova *iova;
164 unsigned int i;
165 int ret;
166
167 info = kzalloc(sizeof(*info), GFP_KERNEL);
168 if (!info)
169 return NULL;
170
171 size = PAGE_ALIGN(size);
172 count = PHYS_PFN(size);
173
> 174 iova = ipu6_alloc_iova(mmu, count);
175 if (!iova)
176 goto out_kfree;
177
178 pages = __alloc_buffer(size, gfp, attrs);
179 if (!pages)
180 goto out_free_iova;
181
182 dev_dbg(dev, "dma_alloc: size %zu iova low pfn %lu, high pfn %lu\n",
183 size, iova->pfn_lo, iova->pfn_hi);
184 for (i = 0; iova->pfn_lo + i <= iova->pfn_hi; i++) {
185 pci_dma_addr = dma_map_page_attrs(&pdev->dev, pages[i], 0,
186 PAGE_SIZE, DMA_BIDIRECTIONAL,
187 attrs);
188 dev_dbg(dev, "dma_alloc: mapped pci_dma_addr %pad\n",
189 &pci_dma_addr);
190 if (dma_mapping_error(&pdev->dev, pci_dma_addr)) {
191 dev_err(dev, "pci_dma_mapping for page[%d] failed", i);
192 goto out_unmap;
193 }
194
195 ret = ipu6_mmu_map(mmu->dmap->mmu_info,
196 PFN_PHYS(iova->pfn_lo + i), pci_dma_addr,
197 PAGE_SIZE);
198 if (ret) {
199 dev_err(dev, "ipu6_mmu_map for pci_dma[%d] %pad failed",
200 i, &pci_dma_addr);
201 dma_unmap_page_attrs(&pdev->dev, pci_dma_addr,
202 PAGE_SIZE, DMA_BIDIRECTIONAL,
203 attrs);
204 goto out_unmap;
205 }
206 }
207
208 info->vaddr = vmap(pages, count, VM_USERMAP, PAGE_KERNEL);
209 if (!info->vaddr)
210 goto out_unmap;
211
212 *dma_handle = PFN_PHYS(iova->pfn_lo);
213
214 info->pages = pages;
215 info->ipu6_iova = *dma_handle;
216 info->size = size;
217 list_add(&info->list, &mmu->vma_list);
218
219 return info->vaddr;
220
221 out_unmap:
222 while (i--) {
223 ipu6_iova = PFN_PHYS(iova->pfn_lo + i);
224 pci_dma_addr = ipu6_mmu_iova_to_phys(mmu->dmap->mmu_info,
225 ipu6_iova);
226 dma_unmap_page_attrs(&pdev->dev, pci_dma_addr, PAGE_SIZE,
227 DMA_BIDIRECTIONAL, attrs);
228
229 ipu6_mmu_unmap(mmu->dmap->mmu_info, ipu6_iova, PAGE_SIZE);
230 }
231
232 __free_buffer(pages, size, attrs);
233
234 out_free_iova:
235 __free_iova(&mmu->dmap->iovad, iova);
236 out_kfree:
237 kfree(info);
238
239 return NULL;
240 }
241 EXPORT_SYMBOL_NS_GPL(ipu6_dma_alloc, "INTEL_IPU6");
242
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] media: intel/ipu6: Fix dma mask for non-secure mode
2025-04-09 9:58 [PATCH] media: intel/ipu6: Fix dma mask for non-secure mode Stanislaw Gruszka
2025-04-09 10:20 ` Stanislaw Gruszka
2025-04-09 12:20 ` kernel test robot
@ 2025-04-09 13:14 ` kernel test robot
2025-04-10 6:53 ` Bingbu Cao
3 siblings, 0 replies; 6+ messages in thread
From: kernel test robot @ 2025-04-09 13:14 UTC (permalink / raw)
To: Stanislaw Gruszka, linux-media
Cc: llvm, oe-kbuild-all, Sakari Ailus, Bingbu Cao
Hi Stanislaw,
kernel test robot noticed the following build errors:
[auto build test ERROR on linuxtv-media-pending/master]
[also build test ERROR on linus/master v6.15-rc1 next-20250409]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]
url: https://github.com/intel-lab-lkp/linux/commits/Stanislaw-Gruszka/media-intel-ipu6-Fix-dma-mask-for-non-secure-mode/20250409-180044
base: https://git.linuxtv.org/media-ci/media-pending.git master
patch link: https://lore.kernel.org/r/20250409095825.1014521-1-stanislaw.gruszka%40linux.intel.com
patch subject: [PATCH] media: intel/ipu6: Fix dma mask for non-secure mode
config: x86_64-buildonly-randconfig-004-20250409 (https://download.01.org/0day-ci/archive/20250409/202504092006.r3kxuxvV-lkp@intel.com/config)
compiler: clang version 20.1.2 (https://github.com/llvm/llvm-project 58df0ef89dd64126512e4ee27b4ac3fd8ddf6247)
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20250409/202504092006.r3kxuxvV-lkp@intel.com/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202504092006.r3kxuxvV-lkp@intel.com/
All errors (new ones prefixed by >>):
>> drivers/media/pci/intel/ipu6/ipu6-dma.c:174:9: error: call to undeclared function 'ipu6_alloc_iova'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
174 | iova = ipu6_alloc_iova(mmu, count);
| ^
drivers/media/pci/intel/ipu6/ipu6-dma.c:174:9: note: did you mean 'ipu_alloc_iova'?
drivers/media/pci/intel/ipu6/ipu6-mmu.h:80:28: note: 'ipu_alloc_iova' declared here
80 | static inline struct iova *ipu_alloc_iova(struct ipu6_mmu *mmu,
| ^
>> drivers/media/pci/intel/ipu6/ipu6-dma.c:174:7: error: incompatible integer to pointer conversion assigning to 'struct iova *' from 'int' [-Wint-conversion]
174 | iova = ipu6_alloc_iova(mmu, count);
| ^ ~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/media/pci/intel/ipu6/ipu6-dma.c:399:9: error: call to undeclared function 'ipu6_alloc_iova'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
399 | iova = ipu6_alloc_iova(mmu, npages);
| ^
drivers/media/pci/intel/ipu6/ipu6-dma.c:399:7: error: incompatible integer to pointer conversion assigning to 'struct iova *' from 'int' [-Wint-conversion]
399 | iova = ipu6_alloc_iova(mmu, npages);
| ^ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
4 errors generated.
--
>> drivers/media/pci/intel/ipu6/ipu6-mmu.c:424:9: error: call to undeclared function 'ipu6_alloc_iova'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
424 | iova = ipu6_alloc_iova(mmu, n_pages);
| ^
drivers/media/pci/intel/ipu6/ipu6-mmu.c:424:9: note: did you mean 'ipu_alloc_iova'?
drivers/media/pci/intel/ipu6/ipu6-mmu.h:80:28: note: 'ipu_alloc_iova' declared here
80 | static inline struct iova *ipu_alloc_iova(struct ipu6_mmu *mmu,
| ^
>> drivers/media/pci/intel/ipu6/ipu6-mmu.c:424:7: error: incompatible integer to pointer conversion assigning to 'struct iova *' from 'int' [-Wint-conversion]
424 | iova = ipu6_alloc_iova(mmu, n_pages);
| ^ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2 errors generated.
vim +/ipu6_alloc_iova +174 drivers/media/pci/intel/ipu6/ipu6-dma.c
151
152 void *ipu6_dma_alloc(struct ipu6_bus_device *sys, size_t size,
153 dma_addr_t *dma_handle, gfp_t gfp,
154 unsigned long attrs)
155 {
156 struct device *dev = &sys->auxdev.dev;
157 struct pci_dev *pdev = sys->isp->pdev;
158 dma_addr_t pci_dma_addr, ipu6_iova;
159 struct ipu6_mmu *mmu = sys->mmu;
160 struct vm_info *info;
161 unsigned long count;
162 struct page **pages;
163 struct iova *iova;
164 unsigned int i;
165 int ret;
166
167 info = kzalloc(sizeof(*info), GFP_KERNEL);
168 if (!info)
169 return NULL;
170
171 size = PAGE_ALIGN(size);
172 count = PHYS_PFN(size);
173
> 174 iova = ipu6_alloc_iova(mmu, count);
175 if (!iova)
176 goto out_kfree;
177
178 pages = __alloc_buffer(size, gfp, attrs);
179 if (!pages)
180 goto out_free_iova;
181
182 dev_dbg(dev, "dma_alloc: size %zu iova low pfn %lu, high pfn %lu\n",
183 size, iova->pfn_lo, iova->pfn_hi);
184 for (i = 0; iova->pfn_lo + i <= iova->pfn_hi; i++) {
185 pci_dma_addr = dma_map_page_attrs(&pdev->dev, pages[i], 0,
186 PAGE_SIZE, DMA_BIDIRECTIONAL,
187 attrs);
188 dev_dbg(dev, "dma_alloc: mapped pci_dma_addr %pad\n",
189 &pci_dma_addr);
190 if (dma_mapping_error(&pdev->dev, pci_dma_addr)) {
191 dev_err(dev, "pci_dma_mapping for page[%d] failed", i);
192 goto out_unmap;
193 }
194
195 ret = ipu6_mmu_map(mmu->dmap->mmu_info,
196 PFN_PHYS(iova->pfn_lo + i), pci_dma_addr,
197 PAGE_SIZE);
198 if (ret) {
199 dev_err(dev, "ipu6_mmu_map for pci_dma[%d] %pad failed",
200 i, &pci_dma_addr);
201 dma_unmap_page_attrs(&pdev->dev, pci_dma_addr,
202 PAGE_SIZE, DMA_BIDIRECTIONAL,
203 attrs);
204 goto out_unmap;
205 }
206 }
207
208 info->vaddr = vmap(pages, count, VM_USERMAP, PAGE_KERNEL);
209 if (!info->vaddr)
210 goto out_unmap;
211
212 *dma_handle = PFN_PHYS(iova->pfn_lo);
213
214 info->pages = pages;
215 info->ipu6_iova = *dma_handle;
216 info->size = size;
217 list_add(&info->list, &mmu->vma_list);
218
219 return info->vaddr;
220
221 out_unmap:
222 while (i--) {
223 ipu6_iova = PFN_PHYS(iova->pfn_lo + i);
224 pci_dma_addr = ipu6_mmu_iova_to_phys(mmu->dmap->mmu_info,
225 ipu6_iova);
226 dma_unmap_page_attrs(&pdev->dev, pci_dma_addr, PAGE_SIZE,
227 DMA_BIDIRECTIONAL, attrs);
228
229 ipu6_mmu_unmap(mmu->dmap->mmu_info, ipu6_iova, PAGE_SIZE);
230 }
231
232 __free_buffer(pages, size, attrs);
233
234 out_free_iova:
235 __free_iova(&mmu->dmap->iovad, iova);
236 out_kfree:
237 kfree(info);
238
239 return NULL;
240 }
241 EXPORT_SYMBOL_NS_GPL(ipu6_dma_alloc, "INTEL_IPU6");
242
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] media: intel/ipu6: Fix dma mask for non-secure mode
2025-04-09 9:58 [PATCH] media: intel/ipu6: Fix dma mask for non-secure mode Stanislaw Gruszka
` (2 preceding siblings ...)
2025-04-09 13:14 ` kernel test robot
@ 2025-04-10 6:53 ` Bingbu Cao
2025-04-10 8:52 ` Stanislaw Gruszka
3 siblings, 1 reply; 6+ messages in thread
From: Bingbu Cao @ 2025-04-10 6:53 UTC (permalink / raw)
To: Stanislaw Gruszka, linux-media; +Cc: Sakari Ailus, Bingbu Cao
Stanislaw,
Thanks for the patch.
On 4/9/25 5:58 PM, Stanislaw Gruszka wrote:
> We use dma_get_mask() of auxdev device for calculate iova pfn limit.
> This is always 32 bit mask as we do not initialize the mask (and we can
> not do so, since dev->dev_mask is NULL anyways for auxdev).
Indeed.
>
> Since we need 31 bit mask for non-secure mode create wrapper of
> alloc_iova() which use mmu_info->aperture_end. This give us always
> the correct mask.
>
> Fixes: daabc5c64703 ("media: ipu6: not override the dma_ops of device in driver")
> Cc: stable@vger.kernel.org
> Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com>
> ---
> drivers/media/pci/intel/ipu6/ipu6-dma.c | 6 ++----
> drivers/media/pci/intel/ipu6/ipu6-dma.h | 7 -------
> drivers/media/pci/intel/ipu6/ipu6-mmu.c | 3 +--
> drivers/media/pci/intel/ipu6/ipu6-mmu.h | 13 +++++++++++++
> 4 files changed, 16 insertions(+), 13 deletions(-)
>
> diff --git a/drivers/media/pci/intel/ipu6/ipu6-dma.c b/drivers/media/pci/intel/ipu6/ipu6-dma.c
> index b34022bad83b..a1d4ec35f802 100644
> --- a/drivers/media/pci/intel/ipu6/ipu6-dma.c
> +++ b/drivers/media/pci/intel/ipu6/ipu6-dma.c
> @@ -171,8 +171,7 @@ void *ipu6_dma_alloc(struct ipu6_bus_device *sys, size_t size,
> size = PAGE_ALIGN(size);
> count = PHYS_PFN(size);
>
> - iova = alloc_iova(&mmu->dmap->iovad, count,
> - PHYS_PFN(dma_get_mask(dev)), 0);
How about directly call?
iova = alloc_iova(&mmu->dmap->iovad, count,
PHYS_PFN(mmu->dmap->mmu_info->aperture_end), 0);
Less change, better.
> + iova = ipu6_alloc_iova(mmu, count);
> if (!iova)
> goto out_kfree;
>
> @@ -397,8 +396,7 @@ int ipu6_dma_map_sg(struct ipu6_bus_device *sys, struct scatterlist *sglist,
> dev_dbg(dev, "dmamap trying to map %d ents %zu pages\n",
> nents, npages);
>
> - iova = alloc_iova(&mmu->dmap->iovad, npages,
> - PHYS_PFN(dma_get_mask(dev)), 0);
> + iova = ipu6_alloc_iova(mmu, npages);
> if (!iova)
> return 0;
>
> diff --git a/drivers/media/pci/intel/ipu6/ipu6-dma.h b/drivers/media/pci/intel/ipu6/ipu6-dma.h
> index b51244add9e6..0dd7c0556bd2 100644
> --- a/drivers/media/pci/intel/ipu6/ipu6-dma.h
> +++ b/drivers/media/pci/intel/ipu6/ipu6-dma.h
> @@ -6,8 +6,6 @@
>
> #include <linux/dma-map-ops.h>
> #include <linux/dma-mapping.h>
> -#include <linux/iova.h>
> -#include <linux/iova.h>
> #include <linux/scatterlist.h>
> #include <linux/types.h>
>
> @@ -15,11 +13,6 @@
>
> struct ipu6_mmu_info;
>
> -struct ipu6_dma_mapping {
> - struct ipu6_mmu_info *mmu_info;
> - struct iova_domain iovad;
> -};
> -
> void ipu6_dma_sync_single(struct ipu6_bus_device *sys, dma_addr_t dma_handle,
> size_t size);
> void ipu6_dma_sync_sg(struct ipu6_bus_device *sys, struct scatterlist *sglist,
> diff --git a/drivers/media/pci/intel/ipu6/ipu6-mmu.c b/drivers/media/pci/intel/ipu6/ipu6-mmu.c
> index 6d1c0b90169d..95eb17855847 100644
> --- a/drivers/media/pci/intel/ipu6/ipu6-mmu.c
> +++ b/drivers/media/pci/intel/ipu6/ipu6-mmu.c
> @@ -421,8 +421,7 @@ static int allocate_trash_buffer(struct ipu6_mmu *mmu)
> int ret;
>
> /* Allocate 8MB in iova range */
> - iova = alloc_iova(&mmu->dmap->iovad, n_pages,
> - PHYS_PFN(mmu->dmap->mmu_info->aperture_end), 0);
> + iova = ipu6_alloc_iova(mmu, n_pages);
> if (!iova) {
> dev_err(mmu->dev, "cannot allocate iova range for trash\n");
> return -ENOMEM;
> diff --git a/drivers/media/pci/intel/ipu6/ipu6-mmu.h b/drivers/media/pci/intel/ipu6/ipu6-mmu.h
> index 8e40b4a66d7d..7f3860796762 100644
> --- a/drivers/media/pci/intel/ipu6/ipu6-mmu.h
> +++ b/drivers/media/pci/intel/ipu6/ipu6-mmu.h
> @@ -7,6 +7,7 @@
> #define ISYS_MMID 1
> #define PSYS_MMID 0
>
> +#include <linux/iova.h>
> #include <linux/list.h>
> #include <linux/spinlock_types.h>
> #include <linux/types.h>
> @@ -58,6 +59,11 @@ struct ipu6_mmu {
> void (*tlb_invalidate)(struct ipu6_mmu *mmu);
> };
>
> +struct ipu6_dma_mapping {
> + struct ipu6_mmu_info *mmu_info;
> + struct iova_domain iovad;
> +};
> +
> struct ipu6_mmu *ipu6_mmu_init(struct device *dev,
> void __iomem *base, int mmid,
> const struct ipu6_hw_variants *hw);
> @@ -70,4 +76,11 @@ void ipu6_mmu_unmap(struct ipu6_mmu_info *mmu_info, unsigned long iova,
> size_t size);
> phys_addr_t ipu6_mmu_iova_to_phys(struct ipu6_mmu_info *mmu_info,
> dma_addr_t iova);
> +
> +static inline struct iova *ipu_alloc_iova(struct ipu6_mmu *mmu,
> + unsigned long n_pages)
> +{
> + return alloc_iova(&mmu->dmap->iovad, n_pages,
> + PHYS_PFN(mmu->dmap->mmu_info->aperture_end), 0);
> +}
> #endif
>
--
Best regards,
Bingbu Cao
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] media: intel/ipu6: Fix dma mask for non-secure mode
2025-04-10 6:53 ` Bingbu Cao
@ 2025-04-10 8:52 ` Stanislaw Gruszka
0 siblings, 0 replies; 6+ messages in thread
From: Stanislaw Gruszka @ 2025-04-10 8:52 UTC (permalink / raw)
To: Bingbu Cao; +Cc: linux-media, Sakari Ailus, Bingbu Cao
Hi Bingbu,
On Thu, Apr 10, 2025 at 02:53:45PM +0800, Bingbu Cao wrote:
> Stanislaw,
>
> Thanks for the patch.
>
> On 4/9/25 5:58 PM, Stanislaw Gruszka wrote:
> > We use dma_get_mask() of auxdev device for calculate iova pfn limit.
> > This is always 32 bit mask as we do not initialize the mask (and we can
> > not do so, since dev->dev_mask is NULL anyways for auxdev).
>
> Indeed.
>
> >
> > Since we need 31 bit mask for non-secure mode create wrapper of
> > alloc_iova() which use mmu_info->aperture_end. This give us always
> > the correct mask.
> >
> > Fixes: daabc5c64703 ("media: ipu6: not override the dma_ops of device in driver")
> > Cc: stable@vger.kernel.org
> > Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com>
> > ---
> > drivers/media/pci/intel/ipu6/ipu6-dma.c | 6 ++----
> > drivers/media/pci/intel/ipu6/ipu6-dma.h | 7 -------
> > drivers/media/pci/intel/ipu6/ipu6-mmu.c | 3 +--
> > drivers/media/pci/intel/ipu6/ipu6-mmu.h | 13 +++++++++++++
> > 4 files changed, 16 insertions(+), 13 deletions(-)
> >
> > diff --git a/drivers/media/pci/intel/ipu6/ipu6-dma.c b/drivers/media/pci/intel/ipu6/ipu6-dma.c
> > index b34022bad83b..a1d4ec35f802 100644
> > --- a/drivers/media/pci/intel/ipu6/ipu6-dma.c
> > +++ b/drivers/media/pci/intel/ipu6/ipu6-dma.c
> > @@ -171,8 +171,7 @@ void *ipu6_dma_alloc(struct ipu6_bus_device *sys, size_t size,
> > size = PAGE_ALIGN(size);
> > count = PHYS_PFN(size);
> >
> > - iova = alloc_iova(&mmu->dmap->iovad, count,
> > - PHYS_PFN(dma_get_mask(dev)), 0);
>
> How about directly call?
> iova = alloc_iova(&mmu->dmap->iovad, count,
> PHYS_PFN(mmu->dmap->mmu_info->aperture_end), 0);
>
> Less change, better.
Yes, direct call results in smaller/nicer fix, I will do this in v2.
Regards
Stanislaw
^ permalink raw reply [flat|nested] 6+ messages in thread