Message ID | 1609861013-129801-2-git-send-email-steven.sistare@oracle.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Series | vfio virtual address update | expand |
Hi Steve, Thank you for the patch! Perhaps something to improve: [auto build test WARNING on vfio/next] [also build test WARNING on v5.11-rc2 next-20210104] [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] url: https://github.com/0day-ci/linux/commits/Steve-Sistare/vfio-virtual-address-update/20210106-000752 base: https://github.com/awilliam/linux-vfio.git next config: x86_64-randconfig-s022-20210105 (attached as .config) compiler: gcc-9 (Debian 9.3.0-15) 9.3.0 reproduce: # apt-get install sparse # sparse version: v0.6.3-208-g46a52ca4-dirty # https://github.com/0day-ci/linux/commit/c7f4454de15f02ea6be7591e8ab4520e21da646e git remote add linux-review https://github.com/0day-ci/linux git fetch --no-tags linux-review Steve-Sistare/vfio-virtual-address-update/20210106-000752 git checkout c7f4454de15f02ea6be7591e8ab4520e21da646e # save the attached .config to linux build tree make W=1 C=1 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__' ARCH=x86_64 If you fix the issue, kindly add following tag as appropriate Reported-by: kernel test robot <lkp@intel.com> "sparse warnings: (new ones prefixed by >>)" >> drivers/vfio/vfio_iommu_type1.c:160:32: sparse: sparse: Using plain integer as NULL pointer drivers/vfio/vfio_iommu_type1.c:176:23: sparse: sparse: Using plain integer as NULL pointer vim +160 drivers/vfio/vfio_iommu_type1.c 147 148 static struct vfio_group *vfio_iommu_find_iommu_group(struct vfio_iommu *iommu, 149 struct iommu_group *iommu_group); 150 151 static void update_pinned_page_dirty_scope(struct vfio_iommu *iommu); 152 /* 153 * This code handles mapping and unmapping of user data buffers 154 * into DMA'ble space using the IOMMU 155 */ 156 157 static struct vfio_dma *vfio_find_dma(struct vfio_iommu *iommu, 158 dma_addr_t start, size_t size) 159 { > 160 struct vfio_dma *res = 0; 161 struct rb_node *node = iommu->dma_list.rb_node; 162 163 while (node) { 164 struct vfio_dma *dma = rb_entry(node, struct vfio_dma, node); 165 166 if (start < dma->iova + dma->size) { 167 res = dma; 168 if (start >= dma->iova) 169 break; 170 node = node->rb_left; 171 } else { 172 node = node->rb_right; 173 } 174 } 175 if (res && size && res->iova >= start + size) 176 res = 0; 177 return res; 178 } 179 --- 0-DAY CI Kernel Test Service, Intel Corporation https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org
Hi Steven, On Tue, 5 Jan 2021 07:36:49 -0800 Steve Sistare <steven.sistare@oracle.com> wrote: > Keep entries properly sorted in the dma_list rb_tree Nothing here changes the order of entries in the tree, they're already sorted. The second chunk is the only thing that touches the tree construction, but that appears to be just a micro optimization that we've already used vfio_find_dma() to verify that a new entry doesn't overlap any existing entries, therefore if the start of the new entry is less than the test entry, the end must also be less. The tree is not changed afaict. > so that iterating > over multiple entries across a range with gaps works, without requiring > one to delete each visited entry as in vfio_dma_do_unmap. As above, I don't see that the tree is changed, so this is just a manipulation of our search function, changing it from a "find any vfio_dma within this range" to a "find the vfio_dma with the lowest iova with this range". But find-any and find-first are computationally different, so I don't think we should blindly replace one with the other. Wouldn't it make more sense to add a vfio_find_first_dma() function in patch 4/ to handle this case? Thanks, Alex > Signed-off-by: Steve Sistare <steven.sistare@oracle.com> > --- > drivers/vfio/vfio_iommu_type1.c | 18 +++++++++++------- > 1 file changed, 11 insertions(+), 7 deletions(-) > > diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c > index 5fbf0c1..02228d0 100644 > --- a/drivers/vfio/vfio_iommu_type1.c > +++ b/drivers/vfio/vfio_iommu_type1.c > @@ -157,20 +157,24 @@ static struct vfio_group *vfio_iommu_find_iommu_group(struct vfio_iommu *iommu, > static struct vfio_dma *vfio_find_dma(struct vfio_iommu *iommu, > dma_addr_t start, size_t size) > { > + struct vfio_dma *res = 0; > struct rb_node *node = iommu->dma_list.rb_node; > > while (node) { > struct vfio_dma *dma = rb_entry(node, struct vfio_dma, node); > > - if (start + size <= dma->iova) > + if (start < dma->iova + dma->size) { > + res = dma; > + if (start >= dma->iova) > + break; > node = node->rb_left; > - else if (start >= dma->iova + dma->size) > + } else { > node = node->rb_right; > - else > - return dma; > + } > } > - > - return NULL; > + if (res && size && res->iova >= start + size) > + res = 0; > + return res; > } > > static void vfio_link_dma(struct vfio_iommu *iommu, struct vfio_dma *new) > @@ -182,7 +186,7 @@ static void vfio_link_dma(struct vfio_iommu *iommu, struct vfio_dma *new) > parent = *link; > dma = rb_entry(parent, struct vfio_dma, node); > > - if (new->iova + new->size <= dma->iova) > + if (new->iova < dma->iova) > link = &(*link)->rb_left; > else > link = &(*link)->rb_right;
On 1/5/2021 7:02 PM, Alex Williamson wrote: > Hi Steven, > > On Tue, 5 Jan 2021 07:36:49 -0800 > Steve Sistare <steven.sistare@oracle.com> wrote: > >> Keep entries properly sorted in the dma_list rb_tree > > Nothing here changes the order of entries in the tree, they're already > sorted. The second chunk is the only thing that touches the tree > construction, but that appears to be just a micro optimization that > we've already used vfio_find_dma() to verify that a new entry doesn't > overlap any existing entries, therefore if the start of the new entry > is less than the test entry, the end must also be less. The tree is > not changed afaict. Agreed. Bad explanation on my part. >> so that iterating >> over multiple entries across a range with gaps works, without requiring >> one to delete each visited entry as in vfio_dma_do_unmap. > > As above, I don't see that the tree is changed, so this is just a > manipulation of our search function, changing it from a "find any > vfio_dma within this range" to a "find the vfio_dma with the lowest > iova with this range". But find-any and find-first are computationally > different, so I don't think we should blindly replace one with the > other. Wouldn't it make more sense to add a vfio_find_first_dma() > function in patch 4/ to handle this case? Thanks, Sure, will do. - Steve > Signed-off-by: Steve Sistare <steven.sistare@oracle.com> >> --- >> drivers/vfio/vfio_iommu_type1.c | 18 +++++++++++------- >> 1 file changed, 11 insertions(+), 7 deletions(-) >> >> diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c >> index 5fbf0c1..02228d0 100644 >> --- a/drivers/vfio/vfio_iommu_type1.c >> +++ b/drivers/vfio/vfio_iommu_type1.c >> @@ -157,20 +157,24 @@ static struct vfio_group *vfio_iommu_find_iommu_group(struct vfio_iommu *iommu, >> static struct vfio_dma *vfio_find_dma(struct vfio_iommu *iommu, >> dma_addr_t start, size_t size) >> { >> + struct vfio_dma *res = 0; >> struct rb_node *node = iommu->dma_list.rb_node; >> >> while (node) { >> struct vfio_dma *dma = rb_entry(node, struct vfio_dma, node); >> >> - if (start + size <= dma->iova) >> + if (start < dma->iova + dma->size) { >> + res = dma; >> + if (start >= dma->iova) >> + break; >> node = node->rb_left; >> - else if (start >= dma->iova + dma->size) >> + } else { >> node = node->rb_right; >> - else >> - return dma; >> + } >> } >> - >> - return NULL; >> + if (res && size && res->iova >= start + size) >> + res = 0; >> + return res; >> } >> >> static void vfio_link_dma(struct vfio_iommu *iommu, struct vfio_dma *new) >> @@ -182,7 +186,7 @@ static void vfio_link_dma(struct vfio_iommu *iommu, struct vfio_dma *new) >> parent = *link; >> dma = rb_entry(parent, struct vfio_dma, node); >> >> - if (new->iova + new->size <= dma->iova) >> + if (new->iova < dma->iova) >> link = &(*link)->rb_left; >> else >> link = &(*link)->rb_right; >
diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c index 5fbf0c1..02228d0 100644 --- a/drivers/vfio/vfio_iommu_type1.c +++ b/drivers/vfio/vfio_iommu_type1.c @@ -157,20 +157,24 @@ static struct vfio_group *vfio_iommu_find_iommu_group(struct vfio_iommu *iommu, static struct vfio_dma *vfio_find_dma(struct vfio_iommu *iommu, dma_addr_t start, size_t size) { + struct vfio_dma *res = 0; struct rb_node *node = iommu->dma_list.rb_node; while (node) { struct vfio_dma *dma = rb_entry(node, struct vfio_dma, node); - if (start + size <= dma->iova) + if (start < dma->iova + dma->size) { + res = dma; + if (start >= dma->iova) + break; node = node->rb_left; - else if (start >= dma->iova + dma->size) + } else { node = node->rb_right; - else - return dma; + } } - - return NULL; + if (res && size && res->iova >= start + size) + res = 0; + return res; } static void vfio_link_dma(struct vfio_iommu *iommu, struct vfio_dma *new) @@ -182,7 +186,7 @@ static void vfio_link_dma(struct vfio_iommu *iommu, struct vfio_dma *new) parent = *link; dma = rb_entry(parent, struct vfio_dma, node); - if (new->iova + new->size <= dma->iova) + if (new->iova < dma->iova) link = &(*link)->rb_left; else link = &(*link)->rb_right;
Keep entries properly sorted in the dma_list rb_tree so that iterating over multiple entries across a range with gaps works, without requiring one to delete each visited entry as in vfio_dma_do_unmap. Signed-off-by: Steve Sistare <steven.sistare@oracle.com> --- drivers/vfio/vfio_iommu_type1.c | 18 +++++++++++------- 1 file changed, 11 insertions(+), 7 deletions(-)