platform/x86: thinkpad_acpi: Revert: Use strndup_user() in dispatch_proc_write()
PATCH 1/1 platform/x86: thinkpad_acpi: Revert: Use strndup_user() in dispatch_proc_write()
Commit 35d13c7a0512 ("platform/x86: thinkpad_acpi: Use strndup_user() in dispatch_proc_write()") cleaned up dispatch_proc_write() by replacing the code to copy the passed in data from userspae with strndup_user().
But strndup_user() expects a 0 terminated input buffer and the buffer passed to dispatch_proc_write() is NOT 0 terminated.
So this change leads to strndup_user() copying some extra random bytes from userspace till it hits a 0 byte.
This commit reverts the change to use strndup_user() fixing the buffer being passed to the ibm_struct.write() call back containing extra junk at the end.
--- drivers/platform/x86/thinkpad_acpi.c | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-)
drivers/platform/x86/thinkpad_acpi.c +11 -3
@@ -886,11 +886,19 @@ static ssize_t dispatch_proc_write(struct file *file,if (!ibm || !ibm->write)return -EINVAL ; +if (count >PAGE_SIZE -1 ) +return -EINVAL ; + + kernbuf =kmalloc (count +1 ,GFP_KERNEL ); +if (!kernbuf) +return -ENOMEM ; - kernbuf =strndup_user (userbuf,PAGE_SIZE ); -if (IS_ERR (kernbuf)) -return PTR_ERR (kernbuf); +if (copy_from_user (kernbuf, userbuf, count)) { +kfree (kernbuf); +return -EFAULT ; + } + kernbuf[count] =0 ; ret = ibm->write (kernbuf);if (ret ==0 ) ret = count;
Signature · 1 line
2.26.2
Quoted 4 lines · Hans de Goede
Commit
35d13c7a0512("platform/x86:thinkpad_acpi: Usestrndup_user()indispatch_proc_write()") cleaned updispatch_proc_write()by replacing the code to copy the passed in data from userspae withstrndup_user().
user space
Quoted 9 lines
But
strndup_user()expects a 0 terminated input buffer and the buffer passed todispatch_proc_write()is NOT 0 terminated.So this change leads to
strndup_user()copying some extra random bytes from userspace till it hits a 0 byte.This commit reverts the change to use
strndup_user()fixing the buffer being passed to theibm_struct.write()call back containing extra junk at the end.
Can we simply use memdup_user()?
And thanks for catching this up!
Signature · 2 lines
With Best Regards, Andy Shevchenko
Quoted 10 lines · Andy Shevchenko
Quoted 4 lines · Hans de Goede
Commit
35d13c7a0512("platform/x86:thinkpad_acpi: Usestrndup_user()indispatch_proc_write()") cleaned updispatch_proc_write()by replacing the code to copy the passed in data from userspae withstrndup_user().user space
Quoted 2 lines
But
strndup_user()expects a 0 terminated input buffer and the buffer passed todispatch_proc_write()is NOT 0 terminated.
Second though, perhaps it's a simple wrong count parameter? strndup_user(..., min(count, PAGE_SIZE)) or so would work?
Quoted 9 lines
Quoted 6 lines
So this change leads to
strndup_user()copying some extra random bytes from userspace till it hits a 0 byte.This commit reverts the change to use
strndup_user()fixing the buffer being passed to theibm_struct.write()call back containing extra junk at the end.Can we simply use
memdup_user()?
And thanks for catching this up!
Signature · 2 lines
With Best Regards, Andy Shevchenko
Hi,
Quoted 9 lines · Andy Shevchenko
Quoted 7 lines · Andy Shevchenko
Quoted 4 lines · Hans de Goede
Commit
35d13c7a0512("platform/x86:thinkpad_acpi: Usestrndup_user()indispatch_proc_write()") cleaned updispatch_proc_write()by replacing the code to copy the passed in data from userspae withstrndup_user().user space
Ack, do you want me to send a v2, or can you fix this up after applying.
Quoted 5 lines
Quoted 2 lines
Quoted 2 lines
But
strndup_user()expects a 0 terminated input buffer and the buffer passed todispatch_proc_write()is NOT 0 terminated.Second though, perhaps it's a simple wrong count parameter?
strndup_user(..., min(count, PAGE_SIZE))or so would work?
I honestly have not looked at a better fix and given that you've just come up with 2 suggestions which may or may not work, combined with where we are in the 5.8 cycle I believe it would be best to just play it safe and go with the revert for 5.8.
If you then take a second attempt at cleaning this up for 5.9 and send it to me, I can test it for you.
Regards,
Hans
Quoted 19 lines · Hans de Goede
Quoted 9 lines · Andy Shevchenko
Quoted 7 lines · Andy Shevchenko
Quoted 4 lines · Hans de Goede
Commit
35d13c7a0512("platform/x86:thinkpad_acpi: Usestrndup_user()indispatch_proc_write()") cleaned updispatch_proc_write()by replacing the code to copy the passed in data from userspae withstrndup_user().user space
Ack, do you want me to send a v2, or can you fix this up after applying.
Quoted 5 lines
Quoted 2 lines
Quoted 2 lines
But
strndup_user()expects a 0 terminated input buffer and the buffer passed todispatch_proc_write()is NOT 0 terminated.Second though, perhaps it's a simple wrong count parameter?
strndup_user(..., min(count, PAGE_SIZE))or so would work?
Quoted 7 lines
I honestly have not looked at a better fix and given that you've just come up with 2 suggestions which may or may not work, combined with where we are in the 5.8 cycle I believe it would be best to just play it safe and go with the revert for 5.8.
If you then take a second attempt at cleaning this up for 5.9 and send it to me, I can test it for you.
I guess there is no need to revert. I have looked at the documentation and see that all of the procfs parameters are normal strings, but you are right they are not NULL terminated per se. So, the problematic place is the size of data we retrieve from user space.
I have sent a patch.
Signature · 2 lines
With Best Regards, Andy Shevchenko