powerpc/powernv/memtrace: Fix out of bounds issue in memtrace mmap
authorRitesh Harjani (IBM) <ritesh.list@gmail.com>
Tue, 10 Jun 2025 02:12:26 +0000 (07:42 +0530)
committerMadhavan Srinivasan <maddy@linux.ibm.com>
Tue, 10 Jun 2025 02:14:27 +0000 (07:44 +0530)
memtrace mmap issue has an out of bounds issue. This patch fixes the by
checking that the requested mapping region size should stay within the
allocated region size.

Reported-by: Jonathan Greental <yonatan02greental@gmail.com>
Fixes: 08a022ad3dfa ("powerpc/powernv/memtrace: Allow mmaping trace buffers")
Signed-off-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Signed-off-by: Madhavan Srinivasan <maddy@linux.ibm.com>
Link: https://patch.msgid.link/20250610021227.361980-1-maddy@linux.ibm.com
arch/powerpc/platforms/powernv/memtrace.c

index 4ac9808e55a44d13a97de03168ce0d95f60d1bb4..2ea30b34335415be91d0aa44ec4f7bf96c4e6d71 100644 (file)
@@ -48,11 +48,15 @@ static ssize_t memtrace_read(struct file *filp, char __user *ubuf,
 static int memtrace_mmap(struct file *filp, struct vm_area_struct *vma)
 {
        struct memtrace_entry *ent = filp->private_data;
+       unsigned long ent_nrpages = ent->size >> PAGE_SHIFT;
+       unsigned long vma_nrpages = vma_pages(vma);
 
-       if (ent->size < vma->vm_end - vma->vm_start)
+       /* The requested page offset should be within object's page count */
+       if (vma->vm_pgoff >= ent_nrpages)
                return -EINVAL;
 
-       if (vma->vm_pgoff << PAGE_SHIFT >= ent->size)
+       /* The requested mapping range should remain within the bounds */
+       if (vma_nrpages > ent_nrpages - vma->vm_pgoff)
                return -EINVAL;
 
        vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);