@@ -141,12 +141,10 @@ iommu_arena_find_pages(struct device *dev, struct pci_iommu_arena *arena,
unsigned long boundary_size;
base = arena->dma_base >> PAGE_SHIFT;
- if (dev) {
- boundary_size = dma_get_seg_boundary(dev) + 1;
- boundary_size >>= PAGE_SHIFT;
- } else {
- boundary_size = 1UL << (32 - PAGE_SHIFT);
- }
+
+ boundary_size = dev ? dma_get_seg_boundary(dev) : U32_MAX;
+ /* Overflow-free shortcut for: ALIGN(b + 1, 1 << s) >> s */
+ boundary_size = (boundary_size >> PAGE_SHIFT) + 1;
/* Search forward for the first mask-aligned sequence of N free ptes */
ptes = arena->ptes;
The boundary_size might be as large as ULONG_MAX, which means that a device has no specific boundary limit. So "+ 1" would potentially overflow. Also, by following other places in the kernel, boundary_size should align with the PAGE_SIZE before right shifting by the PAGE_SHIFT. However, passing it to ALIGN() would potentially overflow too. According to kernel defines: #define ALIGN_MASK(x, mask) (((x) + (mask)) & ~(mask)) #define ALIGN(x, a) ALIGN_MASK(x, (typeof(x))(a) - 1) We can simplify the logic here: ALIGN(boundary + 1, 1 << shift) >> shift = ALIGN_MASK(b + 1, (1 << s) - 1) >> s = {[b + 1 + (1 << s) - 1] & ~[(1 << s) - 1]} >> s = [b + 1 + (1 << s) - 1] >> s = [b + (1 << s)] >> s = (b >> s) + 1 So fixing a potential overflow with the safer shortcut. Signed-off-by: Nicolin Chen <nicoleotsuka@gmail.com> Cc: Christoph Hellwig <hch@lst.de> --- arch/alpha/kernel/pci_iommu.c | 10 ++++------ 1 file changed, 4 insertions(+), 6 deletions(-)