core: Fix undefined behaviour in C11 atomics #11528
Conversation
maribu
added
Area: core
kaspar030
added
CI: ready for build
|
IMO this should do it. Did elf_diff show any differences? |
|
I checked for the Nucleo-F767zi with |
|
This might need a second ACK, too. @SemjonKerner? ;) |
Casting pointers to volatile memory to pointers to regular memory is permitted, but using those pointers to access the memory results in undefined behavior. This commit changes the casts to no longer drop the volatile qualifier. References: https://en.cppreference.com/w/c/language/volatile
| unsigned int mask = irq_disable(); \ | ||
| (void)memmodel; \ | ||
| I##n tmp = *(volatile I##n *)ptr; \ | ||
| *(volatile I##n *)ptr = prefixop(tmp op val); \ |
There was a problem hiding this comment.
weird, github still shows this line's last backslash one off, but with a direct checkout, it is alright?
There was a problem hiding this comment.
On my browser (Firefox) they look all aligned.
I also changed the column of the backslashes as requested to be 4 * n, and not 4 * n + 1. So merge?
There was a problem hiding this comment.
Looks good in github now. You might F5 on this, Kaspar^^
There was a problem hiding this comment.
I F5'ed the shit out of this, still shows the missing space. Anyhow, it also doesn't let me upload the screenshot. Anyways, directly checking this out looks fine, so let's go!
Contribution description
Casting pointers to
volatilememory to pointers to regular memory is permitted, but using those pointers to access the memory results in undefined behavior. This PR changes the casts to no longer drop the volatile qualifier.References: https://en.cppreference.com/w/c/language/volatile
Testing procedure
Compile some applications before and after this PR. The output should not differ.
Issues/PRs references
#11514