lib/test_hmm.c: fix harmless shift wrapping bug
PATCH 1/1 lib/test_hmm.c: fix harmless shift wrapping bug
The "cmd.npages" variable is a u64 that comes from the user. I noticed during review that it could have a shift wrapping bug when it is used in the integer overflow test on the next line.
It turns out this is harmless. The users all do:
unsigned long size = cmd->npages << PAGE_SHIFT;
and after that "size" is used consistently and "cmd->npages" is never used again. So even when there is an integer overflow, everything works fine.
Even though this is harmless, I believe syzbot will complain and fixing it makes the code easier to read.
--- lib/test_hmm.c | 2 ++ 1 file changed, 2 insertions(+)
lib/test_hmm.c +2 -0
@@ -930,6 +930,8 @@ static long dmirror_fops_unlocked_ioctl(struct file *filp,if (cmd.addr & ~PAGE_MASK )return -EINVAL ; +if (cmd.npages >ULONG_MAX >>PAGE_SHIFT ) +return -EINVAL ;if (cmd.addr >= (cmd.addr + (cmd.npages <<PAGE_SHIFT )))return -EINVAL ;
Signature · 1 line
2.30.1
Quoted 26 lines · Dan Carpenter
The "cmd.npages" variable is a u64 that comes from the user. I noticed during review that it could have a shift wrapping bug when it is used in the integer overflow test on the next line.
It turns out this is harmless. The users all do:
unsigned long size = cmd->npages << PAGE_SHIFT;and after that "size" is used consistently and "
cmd->npages" is never used again. So even when there is an integer overflow, everything works fine.Even though this is harmless, I believe syzbot will complain and fixing it makes the code easier to read.
Fixes b2ef9f5a5cb3 ("mm/hmm/test: add selftest driver for HMM")Signed-off-by Dan Carpenter <[email protected]>--- lib/test_hmm.c | 2 ++ 1 file changed, 2 insertions(+)lib/test_hmm.c +0
-0
lib/test_hmm.c static long dmirror_fops_unlocked_ioctl(struct file *filp, · line 936
Quoted 7 lines
lib/test_hmm.c +2
-0@@ -930,6 +930,8 @@ static long dmirror_fops_unlocked_ioctl(struct file *filp,if (cmd.addr & ~PAGE_MASK )return -EINVAL ; +if (cmd.npages >ULONG_MAX >>PAGE_SHIFT ) +return -EINVAL ;if (cmd.addr >= (cmd.addr + (cmd.npages <<PAGE_SHIFT )))return -EINVAL ;
Looks good to me too. Thanks for sending this.