Re: [PATCH RFC 1/2] m68k: Add tracirqs

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi Michael,

On 23/10/2024 09:30, Michael Schmitz wrote:
Hi Jean-Michel,

Am 23.10.2024 um 18:53 schrieb Jean-Michel Hautbois:
Hi Michael,

On 23/10/2024 05:53, Michael Schmitz wrote:
Jean-Michel,

thanks for your patches!

Am 21.10.2024 um 22:44 schrieb Jean-Michel Hautbois:
The TRACE_IRQFLAGS_SUPPORT requires the architecture to call
trace_hardirqs_off() when interrupts are disabled and
trace_hardirqs_on() when they are enabled.
Add those calls to do_IRQ function.

You will also have to add these calls to the three sites in arch/m68k/
coldfire/head.S where interrupts are enabled or disabled.

Thanks for this ! I prepared a v2 with those calls added. I am wondering

Good -

if my second patch is ok, I think it is not, could you please review it

I'll have to read up on how arch_stack_walk is supposed to work - have never seen that code before.

? There may already be something related to the stack in the assembly
part useful to implement arch_stack_walk() in a different way ?

I doubt it


Here:
ENTRY(system_call)
        SAVE_ALL_SYS
        move    #0x2000,%sr             /* enable intrs again */
        GET_CURRENT(%d2)

and here:
ret_from_exception:
        move    #0x2700,%sr             /* disable intrs */
        btst    #5,%sp@(PT_OFF_SR)      /* check if returning to
kernel */
        jeq     Luser_return            /* if so, skip resched,
signals */

and here:
Lwork_to_do:
        movel   %a0@(TINFO_FLAGS),%d1   /* get thread_info->flags */
        move    #0x2000,%sr             /* enable intrs again */
        btst    #TIF_NEED_RESCHED,%d1
        jne     reschedule

There's one similar site in arch/m68k/kernel/head.S
(ret_from_exception) where interrupts are enabled that would need a
similar change, if you want to enable this for all m68k.

I won't be able to test it though ;-).
I see there are a few interrupts disabling in
arch/m68k/include/asm/entry.h too ?

Right - that's different in the coldfire stack save/restore macros. I missed that before.

Looks like wherever SAVE_ALL_SYS, SAVE_ALL_INT and RESTORE_USER are used, calls to trace_irqs_on/off() are required as well.

RESTORE_USER only restores the IPL in the CONFIG_COLDFIRE_SW_A7, which makes this a little tricky ... you might have to add these calls to the macros to get the correct behaviour.


Sadly, not that straightforward.
I have this patch right now:
diff --git a/arch/m68k/Kconfig b/arch/m68k/Kconfig
index b2a3093af677..fc4a2d124514 100644
--- a/arch/m68k/Kconfig
+++ b/arch/m68k/Kconfig
@@ -41,6 +41,7 @@ config M68K
        select ZONE_DMA
        select ARCH_SUPPORTS_RT
        select IRQ_FORCED_THREADING
+       select TRACE_IRQFLAGS_SUPPORT

 config CPU_BIG_ENDIAN
        def_bool y
diff --git a/arch/m68k/coldfire/entry.S b/arch/m68k/coldfire/entry.S
index 4ea08336e2fb..fbdc4404f29e 100644
--- a/arch/m68k/coldfire/entry.S
+++ b/arch/m68k/coldfire/entry.S
@@ -57,6 +57,9 @@ enosys:
 ENTRY(system_call)
        SAVE_ALL_SYS
        move    #0x2000,%sr             /* enable intrs again */
+#ifdef CONFIG_TRACE_IRQFLAGS
+       jbsr    trace_hardirqs_on
+#endif
        GET_CURRENT(%d2)

        cmpl    #NR_syscalls,%d0
@@ -99,6 +102,9 @@ ENTRY(system_call)
        addql   #4,%sp

 ret_from_exception:
+#ifdef CONFIG_TRACE_IRQFLAGS
+       jbsr    trace_hardirqs_off
+#endif
        move    #0x2700,%sr             /* disable intrs */
        btst    #5,%sp@(PT_OFF_SR)      /* check if returning to kernel */
        jeq     Luser_return            /* if so, skip resched, signals */
@@ -140,6 +146,9 @@ Lreturn:
 Lwork_to_do:
        movel   %a0@(TINFO_FLAGS),%d1   /* get thread_info->flags */
        move    #0x2000,%sr             /* enable intrs again */
+#ifdef CONFIG_TRACE_IRQFLAGS
+       jbsr    trace_hardirqs_on
+#endif
        btst    #TIF_NEED_RESCHED,%d1
        jne     reschedule

But it fails when init is called:
[    5.313000] Run /bin/bash as init process
[    5.314000]   with arguments:
[    5.315000]     /bin/bash
[    5.316000]   with environment:
[    5.317000]     HOME=/
[    5.318000]     TERM=linux
[ 5.684000] Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b [ 5.684000] CPU: 0 UID: 0 PID: 1 Comm: bash Not tainted 6.12.0-rc4-00049-g4393ca34ead3 #364
[    5.684000] Stack from 41a03e18:
[ 5.684000] 41a03e18 41540187 41540187 0000000a ffffffff 415a8fb4 4140dd90 41416588 [ 5.684000] 41540187 4140d5de 4102ba84 4100b19c 4100b1a8 00000000 41a38000 4102ba3a [ 5.684000] 41a03ea0 4100c34a 4150e87e 0000000b 0000000b 41a03f80 0000000a 41a3c0d4 [ 5.684000] 41a02000 400004d8 41a08000 4102ba3a 4102ba84 00000000 00000000 000000ff [ 5.684000] 00000000 00000000 41a03ef4 4100cb82 0000000b 0000000b 41a03f80 41a03f6c [ 5.684000] 41016cd2 0000000b 41a03f6c 0000001d 00000026 0000048c ffffffff 00000006
[    5.684000] Call Trace: [<4140dd90>] _printk+0x0/0x18
[    5.684000]  [<41416588>] dump_stack+0xc/0x10
[    5.684000]  [<4140d5de>] panic+0xf2/0x2d4
[    5.684000]  [<4102ba84>] preempt_count_sub+0x0/0x2e
[    5.684000]  [<4100b19c>] arch_local_irq_enable+0x0/0xc
[    5.684000]  [<4100b1a8>] arch_irqs_disabled+0x0/0x10
[    5.684000]  [<4102ba3a>] preempt_count_add+0x0/0x1e
[    5.684000]  [<4100c34a>] do_exit+0x266/0x930
[    5.684000]  [<4102ba3a>] preempt_count_add+0x0/0x1e
[    5.684000]  [<4102ba84>] preempt_count_sub+0x0/0x2e
[    5.684000]  [<4100cb82>] do_group_exit+0x26/0xba
[    5.684000]  [<41016cd2>] get_signal+0x60e/0x76c
[    5.684000]  [<410044e8>] test_ti_thread_flag+0x0/0x14
[    5.684000]  [<4102ba94>] preempt_count_sub+0x10/0x2e
[    5.684000]  [<41004b68>] do_notify_resume+0x3a/0x4c6
[    5.684000]  [<41015d52>] force_sig_fault_to_task+0x32/0x3e
[    5.684000]  [<41015d72>] force_sig_fault+0x14/0x1a
[    5.684000]  [<41005606>] buserr_c+0x9a/0x188
[    5.684000]  [<410065fc>] Lsignal_return+0x14/0x24
[    5.684000]  [<410065de>] Lwork_to_do+0xe/0x18
[    5.684000]
[ 5.684000] ---[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b ]---


JM



Registers %d0-%d5 and %a0-%a2 are saved on the stack at this point and
can be clobbered if need be.

I don't know if they need to be clobbered...

I meant to say that if you need registers to prepare function arguments for trace_irqs_on/off() on the stack, these can be used. But that may not be necessary in this case.

Cheers,

     Michael



Thanks,
JM


Cheers,

     Michael



Signed-off-by: Jean-Michel Hautbois <jeanmichel.hautbois@xxxxxxxxxx>
---
 arch/m68k/Kconfig      | 1 +
 arch/m68k/kernel/irq.c | 2 ++
 2 files changed, 3 insertions(+)

diff --git a/arch/m68k/Kconfig b/arch/m68k/Kconfig
index
cc26df907bfe3c8143a931d259eceabb16af7411..ab3375475721fa63418c40d4ba6ac76679ebc77d
100644
--- a/arch/m68k/Kconfig
+++ b/arch/m68k/Kconfig
@@ -39,6 +39,7 @@ config M68K
     select OLD_SIGSUSPEND3
     select UACCESS_MEMCPY if !MMU
     select ZONE_DMA
+    select TRACE_IRQFLAGS_SUPPORT

 config CPU_BIG_ENDIAN
     def_bool y
diff --git a/arch/m68k/kernel/irq.c b/arch/m68k/kernel/irq.c
index
9ab4f550342e5de11c528f55781432675ffd66bf..74cf60ebbc4bca51f3caa4046dbd2bdb02355711
100644
--- a/arch/m68k/kernel/irq.c
+++ b/arch/m68k/kernel/irq.c
@@ -21,9 +21,11 @@ asmlinkage void do_IRQ(int irq, struct pt_regs *regs)
 {
     struct pt_regs *oldregs = set_irq_regs(regs);

+    trace_hardirqs_off();
     irq_enter();
     generic_handle_irq(irq);
     irq_exit();
+    trace_hardirqs_on();

     set_irq_regs(oldregs);
 }







[Index of Archives]     [Video for Linux]     [Yosemite News]     [Linux S/390]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux