binutils breaks hppa kernel build,
Quoted 12 lines · John David Anglin
Quoted 2 lines
The last "fic" opcode entry is wrong. It's using the wrong instruction format, the mask is wrong, pa10 is wrong, etc.
I don't think we realised this until now. I suspect we've been silently emitting bad code up until this point, and given that it's in
flush_kernel_icache_page, maybe it explains some of our occasional crashes.The code was previously:
fic,m %r23(%r26)
I've committed a fix to binutils head. The above should now be accepted when generating PA 2.0 code. However, I'd probably avoid using this feature for awhile. It looks like the bug has been present since ~ 1996.
Dave
Signature · 6 lines
J. David Anglin [email protected] National Research Council of Canada (613) 990-0752 (FAX: 952-6602) _______________________________________________ parisc-linux mailing list [email protected] http://lists.parisc-linux.org/mailman/listinfo/parisc-linux
Quoted 9 lines · John David Anglin
Quoted 9 lines · John David Anglin
Quoted 2 lines
The last "fic" opcode entry is wrong. It's using the wrong instruction format, the mask is wrong, pa10 is wrong, etc.
I don't think we realised this until now. I suspect we've been silently emitting bad code up until this point, and given that it's in
flush_kernel_icache_page, maybe it explains some of our occasional crashes.
Quoted 8 lines
Quoted 3 lines
The code was previously:
fic,m %r23(%r26)I've committed a fix to binutils head. The above should now be accepted when generating PA 2.0 code. However, I'd probably avoid using this feature for awhile. It looks like the bug has been present since ~ 1996.
This is in a file that's used on pa1.1 processors too, so I've gone with specifying sr4 explicitly. Thanks for fixing this bug and exposing our bug ...
_______________________________________________
parisc-linux mailing list [email protected]
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux