I have to recall my last wrong patch.
In blackfin, kgdb is running in delayed exception IRQ5 other than in
exception IRQ3 directly. Register reti other than retx in pt_regs is
the kgdb return address. So, don't put PC in gdb_regs into retx.
Sonic Zhang
Index: arch/blackfin/kernel/kgdb.c
============================
--- arch/blackfin/kernel/kgdb.c (revision 8105)
+++ arch/blackfin/kernel/kgdb.c (revision 8106)
@@ -147,7 +147,7 @@
regs->lb1 = gdb_regs[BFIN_LB1];
regs->usp = gdb_regs[BFIN_USP];
regs->syscfg = gdb_regs[BFIN_SYSCFG];
- regs->retx = gdb_regs[BFIN_PC];
+ regs->retx = gdb_regs[BFIN_RETX];
regs->retn = gdb_regs[BFIN_RETN];
regs->rete = gdb_regs[BFIN_RETE];
regs->pc = gdb_regs[BFIN_PC];
On Mon, Dec 28, 2009 at 6:17 PM, Zhang, Sonic <Sonic.Zhang@xxxxxxxxxx> wrote:
> A little update.
>
> GDB uses field BFIN_PC as the resume address from debugging mode. But, field
> pc in struct pt_regs is where kernel resumes from an interrupt. So, BFIN_PC
> is actually retx when do kernel debugging.
>
>
> Sonic
>
>
> --- trunk/arch/blackfin/kernel/kgdb.c 2009-12-28 09:27:27 UTC (rev 8085)
> +++ trunk/arch/blackfin/kernel/kgdb.c 2009-12-28 10:08:30 UTC (rev 8086)
> @@ -79,7 +79,7 @@
> gdb_regs[BFIN_RETX] = regs->retx;
> gdb_regs[BFIN_RETN] = regs->retn;
> gdb_regs[BFIN_RETE] = regs->rete;
> - gdb_regs[BFIN_PC] = regs->pc;
> + gdb_regs[BFIN_PC] = regs->retx;
> gdb_regs[BFIN_CC] = 0;
> gdb_regs[BFIN_EXTRA1] = 0;
> gdb_regs[BFIN_EXTRA2] = 0;
> @@ -150,7 +150,7 @@
> regs->retx = gdb_regs[BFIN_PC];
> regs->retn = gdb_regs[BFIN_RETN];
> regs->rete = gdb_regs[BFIN_RETE];
> - regs->pc = gdb_regs[BFIN_PC];
> + regs->pc = gdb_regs[BFIN_RETI];
>
> #if 0 /* can't change these */
> regs->astat = gdb_regs[BFIN_ASTAT];
>
>
>
>
> -----Original Message-----
> From: Mike Frysinger [mailto:vapier.adi@xxxxxxxxx]
> Sent: Sunday, December 27, 2009 5:12 AM
> To: Jason Wessel
> Cc: linux-kernel@xxxxxxxxxxxxxxx; kgdb-bugreport@xxxxxxxxxxxxxxxxxxxxx;
> kdb@xxxxxxxxxxx; mingo@xxxxxxx; Zhang, Sonic
> Subject: Re: [PATCH 09/37] kgdb,blackfin: Add in kgdb_arch_set_pc for blackfin
>
> On Wed, Dec 23, 2009 at 16:19, Jason Wessel wrote:
>> The new debug core api requires all architectures that use to debug
>> core to implement a function to set the program counter.
>>
>> CC: Mike Frysinger <vapier@xxxxxxxxxx>
>> Signed-off-by: Jason Wessel <jason.wessel@xxxxxxxxxxxxx>
>> ---
>> arch/blackfin/kernel/kgdb.c | 5 +++++
>> 1 files changed, 5 insertions(+), 0 deletions(-)
>>
>> diff --git a/arch/blackfin/kernel/kgdb.c b/arch/blackfin/kernel/kgdb.c
>> index f1036b6..5965188 100644
>> --- a/arch/blackfin/kernel/kgdb.c
>> +++ b/arch/blackfin/kernel/kgdb.c
>> @@ -644,6 +644,11 @@ int kgdb_arch_remove_breakpoint(unsigned long
>> addr, char *bundle)
>> return bfin_probe_kernel_write((char *)addr, bundle,
>> BREAK_INSTR_SIZE);
>> }
>>
>> +void kgdb_arch_set_pc(struct pt_regs *regs, unsigned long ip) {
>> + regs->retx = ip;
>> +}
>> +
>> int kgdb_arch_init(void)
>> {
>> kgdb_single_step = 0;
>
> Sonic should be able to check this. our pc handling seems a little wonky atm:
>
> arch/blackfin/kernel/kgdb.c:
> void pt_regs_to_gdb_regs(unsigned long *gdb_regs, struct pt_regs *regs) { ...
> gdb_regs[BFIN_RETI] = regs->pc;
> gdb_regs[BFIN_RETX] = regs->retx;
> gdb_regs[BFIN_PC] = regs->pc;
> ...
> }
> ...
> void gdb_regs_to_pt_regs(unsigned long *gdb_regs, struct pt_regs *regs) { ...
> regs->pc = gdb_regs[BFIN_PC];
> regs->retx = gdb_regs[BFIN_PC];
> /* nothing for BFIN_RETI */
> ...
> }
> -mike
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>
|