Well adding write support is trivial. What I'm more concerned about is if setting page->mapping during allocation of the page could have any negative effect? I of hand don't see any since the page isn't reclaimable directly anyway, but I'm not 100% sure of that. Christian. Am 05.02.2018 um 12:49 schrieb Tom St Denis: > Another thing that occurred to me is this will break write access to > GPU bound memory. Previously we relied on iova to translate the > address and then /dev/mem or /dev/fmem to read/write it. But since > this is literally a read only method obviously there's no write support. > > Tom > > > On 04/02/18 09:56 PM, He, Roger wrote: >> Patch1 & 2 & 4,  Reviewed-by: Roger He <Hongbo.He at amd.com> >> Patch 5: Acked-by: Roger He <Hongbo.He at amd.com> >> >> -----Original Message----- >> From: dri-devel [mailto:dri-devel-bounces at lists.freedesktop.org] On >> Behalf Of Christian K?nig >> Sent: Saturday, February 03, 2018 3:10 AM >> To: amd-gfx at lists.freedesktop.org; dri-devel at lists.freedesktop.org >> Subject: [PATCH 5/5] drm/amdgpu: replace iova debugfs file with iomem >> >> This allows access to pages allocated through the driver with >> optional IOMMU mapping. >> >> Signed-off-by: Christian König <christian.koenig at amd.com> >> --- >>  drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 57 >> ++++++++++++++++++++------------- >>  1 file changed, 35 insertions(+), 22 deletions(-) >> >> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c >> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c >> index 648c449aaa79..795ceaeb82d5 100644 >> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c >> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c >> @@ -1929,38 +1929,51 @@ static const struct file_operations >> amdgpu_ttm_gtt_fops = { >>   #endif >>  -static ssize_t amdgpu_iova_to_phys_read(struct file *f, char >> __user *buf, >> -                  size_t size, loff_t *pos) >> +static ssize_t amdgpu_iomem_read(struct file *f, char __user *buf, >> +                size_t size, loff_t *pos) >>  { >>      struct amdgpu_device *adev = file_inode(f)->i_private; >> -   int r; >> -   uint64_t phys; >>      struct iommu_domain *dom; >> +   ssize_t result = 0; >> +   int r; >>  -   // always return 8 bytes >> -   if (size != 8) >> -       return -EINVAL; >> +   dom = iommu_get_domain_for_dev(adev->dev); >>  -   // only accept page addresses >> -   if (*pos & 0xFFF) >> -       return -EINVAL; >> +   while (size) { >> +       phys_addr_t addr = *pos & PAGE_MASK; >> +       loff_t off = *pos & ~PAGE_MASK; >> +       size_t bytes = PAGE_SIZE - off; >> +       unsigned long pfn; >> +       struct page *p; >> +       void *ptr; >>  -   dom = iommu_get_domain_for_dev(adev->dev); >> -   if (dom) >> -       phys = iommu_iova_to_phys(dom, *pos); >> -   else >> -       phys = *pos; >> +       addr = dom ? iommu_iova_to_phys(dom, addr) : addr; >>  -   r = copy_to_user(buf, &phys, 8); >> -   if (r) >> -       return -EFAULT; >> +       pfn = addr >> PAGE_SHIFT; >> +       if (!pfn_valid(pfn)) >> +           return -EPERM; >> + >> +       p = pfn_to_page(pfn); >> +       if (p->mapping != adev->mman.bdev.dev_mapping) >> +           return -EPERM; >> + >> +       ptr = kmap(p); >> +       r = copy_to_user(buf, ptr, bytes); >> +       kunmap(p); >> +       if (r) >> +           return -EFAULT; >>  -   return 8; >> +       size -= bytes; >> +       *pos += bytes; >> +       result += bytes; >> +   } >> + >> +   return result; >>  } >>  -static const struct file_operations amdgpu_ttm_iova_fops = { >> +static const struct file_operations amdgpu_ttm_iomem_fops = { >>      .owner = THIS_MODULE, >> -   .read = amdgpu_iova_to_phys_read, >> +   .read = amdgpu_iomem_read, >>      .llseek = default_llseek >>  }; >>  @@ -1973,7 +1986,7 @@ static const struct { #ifdef >> CONFIG_DRM_AMDGPU_GART_DEBUGFS >>      { "amdgpu_gtt", &amdgpu_ttm_gtt_fops, TTM_PL_TT }, #endif >> -   { "amdgpu_iova", &amdgpu_ttm_iova_fops, TTM_PL_SYSTEM }, >> +   { "amdgpu_iomem", &amdgpu_ttm_iomem_fops, TTM_PL_SYSTEM }, >>  }; >>   #endif >> -- >> 2.14.1 >> >> _______________________________________________ >> dri-devel mailing list >> dri-devel at lists.freedesktop.org >> https://lists.freedesktop.org/mailman/listinfo/dri-devel >> _______________________________________________ >> amd-gfx mailing list >> amd-gfx at lists.freedesktop.org >> https://lists.freedesktop.org/mailman/listinfo/amd-gfx >> >
