gcc3 compatibility and debugger support
A message that is not in the archive (sent off-list or not yet received): [email protected]
Great Franz!
I could rebuild working kernel(2.4 devel) with gcc-3.0.2 pre
Thanks
Kaoru
From: Franz Sirl <[email protected]>
To: [email protected], [email protected]
Subject: [PATCH] gcc3 compatibility and debugger support
Date: Thu, 6 Sep 2001 21:32:59 +0200
Quoted 11 lines
Hi,
this patch fixes up the RELOC macro (which relies on implentation defined things in C), so it's possible again to boot a gcc3 compiled kernel. The patch guards the problematic code with a NOP inline asm, preventing the problematic (but legal) gcc3 optimization.
The second part makes sure that the current task ends up in a known memory location even for UP. This helps JTAG/BDM debuggers a lot (it's very expensive to stop CPU/peek CPU reg/restart CPU).The patch is against 2_4. In 2_4_devel the gcc3 workaround in
prom.ccan
be
Quoted 3 lines
removed too.
Franz.
** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/
Review
Quoted 4 lines · Franz Sirl
(quoted diff) +2
-0current_set[smp_processor_id()] = new; +#else + current_set[0] = new; #endif /* CONFIG_SMP */
Yuck. Just put the #endif before the line that sets current_set[]. smp_processor_id() is 0 on UP.
Paul.
** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/
Quoted 8 lines · Paul Mackerras
Review
Quoted 4 lines · Franz Sirl
(quoted diff) +2
-0current_set[smp_processor_id()] = new; +#else + current_set[0] = new; #endif /* CONFIG_SMP */Yuck. Just put the #endif before the line that sets current_set[].
smp_processor_id()is 0 on UP.
OK, here is the updated patch.
Franz.
===== arch/ppc/kernel/process.c 1.29 vs edited =====
arch/ppc/kernel/process.c +3 -1
@@ -226,8 +226,10 @@if ((prev->thread.regs && (prev->thread.regs->msr &MSR_VEC )))giveup_altivec (prev);#endif /* CONFIG_ALTIVEC */ - current_set[smp_processor_id ()] = new;#endif /* CONFIG_SMP */ + + current_set[smp_processor_id ()] = new; +/* Avoid the trap. On smp this this never happens since * we don't set last_task_used_altivec -- Cort */
===== include/asm-ppc/prom.h 1.20 vs edited =====
include/asm-ppc/prom.h +1 -1
@@ -106,7 +106,7 @@#define PTRRELOC (x) ((typeof(x))((unsigned long)(x) + offset))#define PTRUNRELOC (x) ((typeof(x))((unsigned long)(x) - offset)) -#define RELOC (x) (*PTRRELOC(&(x))) +#define RELOC (x) (*({ typeof(x) * __ptr = PTRRELOC(&(x)); __asm__ ("" : "=r" (__ptr) : "0" (__ptr)); __ptr;}))#endif /* _PPC_PROM_H */ #endif /* __KERNEL__ */