| Message ID | 20200202171635.4039044-1-chris@chris-wilson.co.uk (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [1/5] drm: Remove PageReserved manipulation from drm_pci_alloc | expand |
Hi Chris. On Sun, Feb 02, 2020 at 05:16:31PM +0000, Chris Wilson wrote: > drm_pci_alloc/drm_pci_free are very thin wrappers around the core dma > facilities, and we have no special reason within the drm layer to behave > differently. In particular, since > > commit de09d31dd38a50fdce106c15abd68432eebbd014 > Author: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> > Date: Fri Jan 15 16:51:42 2016 -0800 > > page-flags: define PG_reserved behavior on compound pages > > As far as I can see there's no users of PG_reserved on compound pages. > Let's use PF_NO_COMPOUND here. > > it has been illegal to combine GFP_COMP with SetPageReserved, so lets > stop doing both and leave the dma layer to its own devices. > > Reported-by: Taketo Kabe > Closes: https://gitlab.freedesktop.org/drm/intel/issues/1027 > Fixes: de09d31dd38a ("page-flags: define PG_reserved behavior on compound pages") > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> > Cc: <stable@vger.kernel.org> # v4.5+ > --- Out of curiosity I added the full series and tried to build it. I did not really take the time to review the patches. It failed with alpha allmodconfig like this: (On top of drm-misc-next as of yesterday) /home/sam/drm/linux.git/drivers/gpu/drm/r128/ati_pcigart.c: In function ‘drm_ati_alloc_pcigart_table’: /home/sam/drm/linux.git/drivers/gpu/drm/r128/ati_pcigart.c:50:7: error: expected expression before ‘^’ token 50 | ^gart_info->bus_addr, | ^ /home/sam/drm/linux.git/drivers/gpu/drm/r128/ati_pcigart.c:48:3: error: too few arguments to function ‘dma_alloc_coherent’ 48 | dma_alloc_coherent(&dev->pdev->dev, | ^~~~~~~~~~~~~~~~~~ In file included from /home/sam/drm/linux.git/arch/alpha/include/asm/pci.h:8, from /home/sam/drm/linux.git/include/linux/pci.h:1777, from /home/sam/drm/linux.git/include/drm/drm_pci.h:35, from /home/sam/drm/linux.git/drivers/gpu/drm/r128/ati_pcigart.c:37: /home/sam/drm/linux.git/include/linux/dma-mapping.h:641:21: note: declared here 641 | static inline void *dma_alloc_coherent(struct device *dev, size_t size, | ^~~~~~~~~~~~~~~~~~ /home/sam/drm/linux.git/drivers/gpu/drm/r128/ati_pcigart.c: At top level: /home/sam/drm/linux.git/drivers/gpu/drm/r128/ati_pcigart.c:101:2: error: expected identifier or ‘(’ before ‘return’ 101 | return 1; | ^~~~~~ /home/sam/drm/linux.git/drivers/gpu/drm/r128/ati_pcigart.c:102:1: error: expected identifier or ‘(’ before ‘}’ token 102 | } | ^ /home/sam/drm/linux.git/drivers/gpu/drm/r128/ati_pcigart.c: In function ‘drm_ati_pcigart_init’: /home/sam/drm/linux.git/drivers/gpu/drm/r128/ati_pcigart.c:168:13: warning: assignment to ‘u32 *’ {aka ‘unsigned int *’} from ‘unsigned int’ makes pointer from integer without a cast [-Wint-conversion] 168 | page_base = (u32) entry->busaddr[i]; | ^ /home/sam/drm/linux.git/drivers/gpu/drm/r128/ati_pcigart.c:176:21: error: invalid operands to binary | (have ‘u32 *’ {aka ‘unsigned int *’} and ‘int’) 176 | val = page_base | 0xc; | ^ /home/sam/drm/linux.git/drivers/gpu/drm/r128/ati_pcigart.c:179:22: error: invalid operands to binary >> (have ‘u32 *’ {aka ‘unsigned int *’} and ‘int’) 179 | val = (page_base >> 8) | 0xc; | ^~ /home/sam/drm/linux.git/drivers/gpu/drm/r128/ati_pcigart.c:183:9: warning: assignment to ‘u32’ {aka ‘unsigned int’} from ‘u32 *’ {aka ‘unsigned int *’} makes integer from pointer without a cast [-Wint-conversion] 183 | val = page_base; | ^ /home/sam/drm/linux.git/drivers/gpu/drm/r128/ati_pcigart.c:188:12: warning: dereferencing ‘void *’ pointer 188 | address[gart_idx] = cpu_to_le32(val); I did not try other architectures and did not try to fix it. When I applied the patches checkpatch was no too happy: Applying: drm: Remove PageReserved manipulation from drm_pci_alloc [sam-pci fa4a1146af59] drm: Remove PageReserved manipulation from drm_pci_alloc Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Sun Feb 2 17:16:31 2020 +0000 1 file changed, 2 insertions(+), 21 deletions(-) fa4a1146af59 (HEAD -> sam-pci) drm: Remove PageReserved manipulation from drm_pci_alloc -:10: ERROR:GIT_COMMIT_ID: Please use git commit description style 'commit <12+ chars of sha1> ("<title line>")' - ie: 'commit de09d31dd38a ("page-flags: define PG_reserved behavior on compound pages")' #10: commit de09d31dd38a50fdce106c15abd68432eebbd014 -:22: ERROR:BAD_SIGN_OFF: Unrecognized email address: 'Taketo Kabe' #22: Reported-by: Taketo Kabe [sam-pci b553cf48b548] drm/r128: Wean off drm_pci_alloc Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Sun Feb 2 17:16:33 2020 +0000 2 files changed, 17 insertions(+), 17 deletions(-) b553cf48b548 (HEAD -> sam-pci) drm/r128: Wean off drm_pci_alloc -:13: WARNING:OBSOLETE: drivers/gpu/drm/r128/ati_pcigart.c is marked as 'obsolete' in the MAINTAINERS hierarchy. No unnecessary modifications please. -:16: WARNING:OBSOLETE: drivers/gpu/drm/r128/ati_pcigart.c is marked as 'obsolete' in the MAINTAINERS hierarchy. No unnecessary modifications please. -:26: CHECK:PARENTHESIS_ALIGNMENT: Alignment should match open parenthesis #26: FILE: drivers/gpu/drm/r128/ati_pcigart.c:49: + dma_alloc_coherent(&dev->pdev->dev, + gart_info->table_size, -:27: CHECK:SPACING: spaces preferred around that '^' (ctx:ExV) #27: FILE: drivers/gpu/drm/r128/ati_pcigart.c:50: + ^gart_info->bus_addr, ^ [sam-pci 7e77c3ee282d] drm: Remove exports for drm_pci_alloc/drm_pci_free Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Sun Feb 2 17:16:35 2020 +0000 4 files changed, 32 insertions(+), 45 deletions(-) 7e77c3ee282d (HEAD -> sam-pci) drm: Remove exports for drm_pci_alloc/drm_pci_free -:53: CHECK:LINE_SPACING: Please don't use multiple blank lines #53: FILE: drivers/gpu/drm/drm_legacy.h:215: + -:58: ERROR:POINTER_LOCATION: "foo * bar" should be "foo *bar" #58: FILE: drivers/gpu/drm/drm_legacy.h:220: +void drm_legacy_pci_free(struct drm_device *dev, struct drm_dma_handle * dmah); -:100: ERROR:POINTER_LOCATION: "foo * bar" should be "foo *bar" #100: FILE: drivers/gpu/drm/drm_pci.c:42: +drm_legacy_pci_alloc(struct drm_device * dev, size_t size, size_t align) -:119: ERROR:POINTER_LOCATION: "foo * bar" should be "foo *bar" #119: FILE: drivers/gpu/drm/drm_pci.c:70: +void drm_legacy_pci_free(struct drm_device * dev, drm_dma_handle_t * dmah) -:119: ERROR:POINTER_LOCATION: "foo * bar" should be "foo *bar" #119: FILE: drivers/gpu/drm/drm_pci.c:70: +void drm_legacy_pci_free(struct drm_device * dev, drm_dma_handle_t * dmah) Sam
On Sun, Feb 2, 2020 at 12:16 PM Chris Wilson <chris@chris-wilson.co.uk> wrote: > > drm_pci_alloc/drm_pci_free are very thin wrappers around the core dma > facilities, and we have no special reason within the drm layer to behave > differently. In particular, since > > commit de09d31dd38a50fdce106c15abd68432eebbd014 > Author: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> > Date: Fri Jan 15 16:51:42 2016 -0800 > > page-flags: define PG_reserved behavior on compound pages > > As far as I can see there's no users of PG_reserved on compound pages. > Let's use PF_NO_COMPOUND here. > > it has been illegal to combine GFP_COMP with SetPageReserved, so lets > stop doing both and leave the dma layer to its own devices. > > Reported-by: Taketo Kabe Needs an email address. > Closes: https://gitlab.freedesktop.org/drm/intel/issues/1027 Should be Bug: rather than Closes: > Fixes: de09d31dd38a ("page-flags: define PG_reserved behavior on compound pages") > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> > Cc: <stable@vger.kernel.org> # v4.5+ With those fixed: Reviewed-by: Alex Deucher <alexander.deucher@amd.com> > --- > drivers/gpu/drm/drm_pci.c | 23 ++--------------------- > 1 file changed, 2 insertions(+), 21 deletions(-) > > diff --git a/drivers/gpu/drm/drm_pci.c b/drivers/gpu/drm/drm_pci.c > index f2e43d341980..d16dac4325f9 100644 > --- a/drivers/gpu/drm/drm_pci.c > +++ b/drivers/gpu/drm/drm_pci.c > @@ -51,8 +51,6 @@ > drm_dma_handle_t *drm_pci_alloc(struct drm_device * dev, size_t size, size_t align) > { > drm_dma_handle_t *dmah; > - unsigned long addr; > - size_t sz; > > /* pci_alloc_consistent only guarantees alignment to the smallest > * PAGE_SIZE order which is greater than or equal to the requested size. > @@ -68,20 +66,13 @@ drm_dma_handle_t *drm_pci_alloc(struct drm_device * dev, size_t size, size_t ali > dmah->size = size; > dmah->vaddr = dma_alloc_coherent(&dev->pdev->dev, size, > &dmah->busaddr, > - GFP_KERNEL | __GFP_COMP); > + GFP_KERNEL); > > if (dmah->vaddr == NULL) { > kfree(dmah); > return NULL; > } > > - /* XXX - Is virt_to_page() legal for consistent mem? */ > - /* Reserve */ > - for (addr = (unsigned long)dmah->vaddr, sz = size; > - sz > 0; addr += PAGE_SIZE, sz -= PAGE_SIZE) { > - SetPageReserved(virt_to_page((void *)addr)); > - } > - > return dmah; > } > > @@ -94,19 +85,9 @@ EXPORT_SYMBOL(drm_pci_alloc); > */ > void __drm_legacy_pci_free(struct drm_device * dev, drm_dma_handle_t * dmah) > { > - unsigned long addr; > - size_t sz; > - > - if (dmah->vaddr) { > - /* XXX - Is virt_to_page() legal for consistent mem? */ > - /* Unreserve */ > - for (addr = (unsigned long)dmah->vaddr, sz = dmah->size; > - sz > 0; addr += PAGE_SIZE, sz -= PAGE_SIZE) { > - ClearPageReserved(virt_to_page((void *)addr)); > - } > + if (dmah->vaddr) > dma_free_coherent(&dev->pdev->dev, dmah->size, dmah->vaddr, > dmah->busaddr); > - } > } > > /** > -- > 2.25.0 > > _______________________________________________ > dri-devel mailing list > dri-devel@lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/dri-devel
Quoting Alex Deucher (2020-02-03 21:49:48) > On Sun, Feb 2, 2020 at 12:16 PM Chris Wilson <chris@chris-wilson.co.uk> wrote: > > > > drm_pci_alloc/drm_pci_free are very thin wrappers around the core dma > > facilities, and we have no special reason within the drm layer to behave > > differently. In particular, since > > > > commit de09d31dd38a50fdce106c15abd68432eebbd014 > > Author: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> > > Date: Fri Jan 15 16:51:42 2016 -0800 > > > > page-flags: define PG_reserved behavior on compound pages > > > > As far as I can see there's no users of PG_reserved on compound pages. > > Let's use PF_NO_COMPOUND here. > > > > it has been illegal to combine GFP_COMP with SetPageReserved, so lets > > stop doing both and leave the dma layer to its own devices. > > > > Reported-by: Taketo Kabe > > Needs an email address. None provided, I don't insist that they opt in to potential spam harvesting. > > Closes: https://gitlab.freedesktop.org/drm/intel/issues/1027 > > Should be Bug: rather than Closes: We're using Closes for gitlab, since we hope to integrate with gitlab someday. (Or at least some integrated bug/source management, of which gitlab is the current forerunner.) -Chris
diff --git a/drivers/gpu/drm/drm_pci.c b/drivers/gpu/drm/drm_pci.c index f2e43d341980..d16dac4325f9 100644 --- a/drivers/gpu/drm/drm_pci.c +++ b/drivers/gpu/drm/drm_pci.c @@ -51,8 +51,6 @@ drm_dma_handle_t *drm_pci_alloc(struct drm_device * dev, size_t size, size_t align) { drm_dma_handle_t *dmah; - unsigned long addr; - size_t sz; /* pci_alloc_consistent only guarantees alignment to the smallest * PAGE_SIZE order which is greater than or equal to the requested size. @@ -68,20 +66,13 @@ drm_dma_handle_t *drm_pci_alloc(struct drm_device * dev, size_t size, size_t ali dmah->size = size; dmah->vaddr = dma_alloc_coherent(&dev->pdev->dev, size, &dmah->busaddr, - GFP_KERNEL | __GFP_COMP); + GFP_KERNEL); if (dmah->vaddr == NULL) { kfree(dmah); return NULL; } - /* XXX - Is virt_to_page() legal for consistent mem? */ - /* Reserve */ - for (addr = (unsigned long)dmah->vaddr, sz = size; - sz > 0; addr += PAGE_SIZE, sz -= PAGE_SIZE) { - SetPageReserved(virt_to_page((void *)addr)); - } - return dmah; } @@ -94,19 +85,9 @@ EXPORT_SYMBOL(drm_pci_alloc); */ void __drm_legacy_pci_free(struct drm_device * dev, drm_dma_handle_t * dmah) { - unsigned long addr; - size_t sz; - - if (dmah->vaddr) { - /* XXX - Is virt_to_page() legal for consistent mem? */ - /* Unreserve */ - for (addr = (unsigned long)dmah->vaddr, sz = dmah->size; - sz > 0; addr += PAGE_SIZE, sz -= PAGE_SIZE) { - ClearPageReserved(virt_to_page((void *)addr)); - } + if (dmah->vaddr) dma_free_coherent(&dev->pdev->dev, dmah->size, dmah->vaddr, dmah->busaddr); - } } /**
drm_pci_alloc/drm_pci_free are very thin wrappers around the core dma facilities, and we have no special reason within the drm layer to behave differently. In particular, since commit de09d31dd38a50fdce106c15abd68432eebbd014 Author: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Date: Fri Jan 15 16:51:42 2016 -0800 page-flags: define PG_reserved behavior on compound pages As far as I can see there's no users of PG_reserved on compound pages. Let's use PF_NO_COMPOUND here. it has been illegal to combine GFP_COMP with SetPageReserved, so lets stop doing both and leave the dma layer to its own devices. Reported-by: Taketo Kabe Closes: https://gitlab.freedesktop.org/drm/intel/issues/1027 Fixes: de09d31dd38a ("page-flags: define PG_reserved behavior on compound pages") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: <stable@vger.kernel.org> # v4.5+ --- drivers/gpu/drm/drm_pci.c | 23 ++--------------------- 1 file changed, 2 insertions(+), 21 deletions(-)