https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81715
--- Comment #6 from Arnd Bergmann <arnd at linaro dot org> --- (In reply to Martin Liška from comment #5) > I can confirm that for the biggest function 'nl80211_send_wiphy', it really > contains majority of stack variables which are 4B large. Having an adaptive > redzone mechanism, one can effectively save half of stack. That will still > not be enough for 1024B. To clarify the limit: The test output was from a 64-bit build, which currently has a 2048 byte limit in the kernel, while 32-bit targets use a 1024 byte limit. My series that avoids the problem here and in other files by adding optional 'noinline' attributes has a follow-up series that lowers the 64-bit limit to 1536 bytes after addressing all the warnings, but that is not currently posted for inclusion. If we can get gcc-8 to stay below 2048 bytes for all files without the noinline annotation, we could still keep using that limit. > Anyway, I there's another example where one needs to limit (or disable ASAN) > in kernel: https://patchwork.kernel.org/patch/9653417/ If I remember correctly, that one only happens with -fsanitize-address-use-after-scope, and I had a workaround that avoids this by changing the kernel code.
