With LInux 5.6 commit 47a2ebb7f5 a new warning from the kernel has been added of the form:
process '/usr/bin/hdhomerun_record' started with executable stack
The kernel git log documents some cases where code practices can result in this artifact (and, as I recall, uclibc has a configuration option regarding executable stacks).
If possible, not having an executable stack is typically considered desirable, even though without address space randomization it has limited benefits.
Thanks for the consideration.
Linux 5.6 new warning for hdhomerun_record regarding executable stack
Re: Linux 5.6 new warning for hdhomerun_record regarding executable stack
I ran a series of checks and fixed one asm file in our code base which wasn't configured to indicate non-executable stack.
Unfortunately it looks like we will have to upgrade and recompile the toolchain to fix the gcc library... will look into that tomorrow.
Nick
Unfortunately it looks like we will have to upgrade and recompile the toolchain to fix the gcc library... will look into that tomorrow.
Nick
Re: Linux 5.6 new warning for hdhomerun_record regarding executable stack
Ouch. Upgrading an entire tool-chain can require a lot of re-validation and certification in the embedded device world (and is why not so long ago I was aware of orgs stuck on things like gcc 2.95).
I did not really expect this to get attention for a while, so thanks for looking at this so quickly.