perf: Apply e9a6882f267a to v4.9 stable (fix for null-pointer dereference)

perf: Apply e9a6882f267a to v4.9 stable (fix for null-pointer dereference)

[Alexander Dahl]

Hei hei,

please apply e9a6882f267a8105461066e3ea6b4b6b9be1b807 ("perf event: Check 
ref_reloc_sym before using it") to the v4.9 stable tree and maybe to later LTS 
trees (v4.14, v4.19) as well. 

That change is in mainline since v5.4-rc1 and it's a fix for a possible null 
pointer dereference in the tool 'perf', so it is a fix for a user space tool 
coming with the kernel tree actually.

I am directly affeted by this, calling 'perf record' on an at91 based device 
(armv5te, sam9g20 soc) running PREEMPT RT kernel 4.9.220-rt143 results in a 
segfault here.  I debugged this with gdb and tracked it down to the exact 
pointer being NULL, which is checked now in that above mentioned changeset.

I could cleanly apply that changeset to my local tree and the segfault is not 
triggered anymore, I can use perf on that platform now.

I only tested this on the above mentioned v4.9 based kernel version.

Thanks & Greets
Alex

Re: perf: Apply e9a6882f267a to v4.9 stable (fix for null-pointer dereference)

[Greg KH]

On Wed, Nov 25, 2020 at 11:01:39AM +0100, Alexander Dahl wrote:
> Hei hei,
> 
> please apply e9a6882f267a8105461066e3ea6b4b6b9be1b807 ("perf event: Check 
> ref_reloc_sym before using it") to the v4.9 stable tree and maybe to later LTS 
> trees (v4.14, v4.19) as well. 
> 
> That change is in mainline since v5.4-rc1 and it's a fix for a possible null 
> pointer dereference in the tool 'perf', so it is a fix for a user space tool 
> coming with the kernel tree actually.
> 
> I am directly affeted by this, calling 'perf record' on an at91 based device 
> (armv5te, sam9g20 soc) running PREEMPT RT kernel 4.9.220-rt143 results in a 
> segfault here.  I debugged this with gdb and tracked it down to the exact 
> pointer being NULL, which is checked now in that above mentioned changeset.
> 
> I could cleanly apply that changeset to my local tree and the segfault is not 
> triggered anymore, I can use perf on that platform now.
> 
> I only tested this on the above mentioned v4.9 based kernel version.

Now queued up, thanks.

greg k-h