linux-kernel - Re: [RFC V2 7/8] powerpc: Enable IRQ generic entry/exit path.

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <6a6de5ca-d105-466b-b44a-8bf084f2a924@linux.ibm.com>
Date: Thu, 18 Sep 2025 12:25:55 +0530
From: Mukesh Kumar Chaurasiya <mkchauras@...ux.ibm.com>
To: Shrikanth Hegde <sshegde@...ux.ibm.com>,
        Mukesh Kumar Chaurasiya <mchauras@...ux.ibm.com>, maddy@...ux.ibm.com,
        mpe@...erman.id.au, npiggin@...il.com, christophe.leroy@...roup.eu
Cc: oleg@...hat.com, kees@...nel.org, luto@...capital.net, wad@...omium.org,
        deller@....de, ldv@...ace.io, macro@...am.me.uk, charlie@...osinc.com,
        akpm@...ux-foundation.org, bigeasy@...utronix.de,
        ankur.a.arora@...cle.com, naveen@...nel.org,
        thomas.weissschuh@...utronix.de, Jason@...c4.com, peterz@...radead.org,
        tglx@...utronix.de, namcao@...utronix.de, kan.liang@...ux.intel.com,
        mingo@...nel.org, oliver.upton@...ux.dev, mark.barnett@....com,
        atrajeev@...ux.vnet.ibm.com, rppt@...nel.org, coltonlewis@...gle.com,
        linuxppc-dev@...ts.ozlabs.org, linux-kernel@...r.kernel.org
Subject: Re: [RFC V2 7/8] powerpc: Enable IRQ generic entry/exit path.


On 9/16/25 9:46 AM, Shrikanth Hegde wrote:
>
>
> On 9/9/25 2:32 AM, Mukesh Kumar Chaurasiya wrote:
>> Enable generic entry/exit path for ppc irq.
>>
>> Signed-off-by: Mukesh Kumar Chaurasiya <mchauras@...ux.ibm.com>
>> ---
>>   arch/powerpc/Kconfig                    |   1 +
>>   arch/powerpc/include/asm/entry-common.h |  93 ++---
>>   arch/powerpc/include/asm/interrupt.h    | 492 +++---------------------
>>   arch/powerpc/kernel/interrupt.c         |   9 +-
>>   arch/powerpc/kernel/interrupt_64.S      |   2 -
>>   5 files changed, 92 insertions(+), 505 deletions(-)
>>
>                                 \
>> diff --git a/arch/powerpc/kernel/interrupt.c 
>> b/arch/powerpc/kernel/interrupt.c
>> index f53d432f60870..7bb8a31b24ea7 100644
>> --- a/arch/powerpc/kernel/interrupt.c
>> +++ b/arch/powerpc/kernel/interrupt.c
>> @@ -297,13 +297,8 @@ notrace unsigned long 
>> interrupt_exit_kernel_prepare(struct pt_regs *regs)
>>           /* Returning to a kernel context with local irqs enabled. */
>>           WARN_ON_ONCE(!(regs->msr & MSR_EE));
>>   again:
>> -        if (need_irq_preemption()) {
>> -            /* Return to preemptible kernel context */
>> -            if (unlikely(read_thread_flags() & _TIF_NEED_RESCHED)) {
>> -                if (preempt_count() == 0)
>> -                    preempt_schedule_irq();
>> -            }
>> -        }
>> +        if (need_irq_preemption())
>> +            irqentry_exit_cond_resched();
>
> irqentry_exit_cond_resched is also called in irqentry_exit. It would 
> be better if we can find ways to avoid calling it again.
>
> I see a loop here. But comment says it is not enabling irq again. so 
> the loop is bounded. So might be okay to remove cond_resched here. do run
> preemptirq, irq tracers to ensure that is case.
>
Sure.
> Also, what is this "soft_interrupts"?

You mean soft masked interrupts?
It's a mechanism to buffer interrupts without disabling the ee bit so 
that we can replay those interrupts later.

>>             check_return_regs_valid(regs);
>>   diff --git a/arch/powerpc/kernel/interrupt_64.S 
>> b/arch/powerpc/kernel/interrupt_64.S
>> index 1ad059a9e2fef..6aa88fe91fb6a 100644
>> --- a/arch/powerpc/kernel/interrupt_64.S
>> +++ b/arch/powerpc/kernel/interrupt_64.S
>> @@ -418,8 +418,6 @@ _ASM_NOKPROBE_SYMBOL(interrupt_return_\srr\())
>>       beq    interrupt_return_\srr\()_kernel
>>   interrupt_return_\srr\()_user: /* make backtraces match the _kernel 
>> variant */
>>   _ASM_NOKPROBE_SYMBOL(interrupt_return_\srr\()_user)
>> -    addi    r3,r1,STACK_INT_FRAME_REGS
>> -    bl    CFUNC(interrupt_exit_user_prepare)
>>   #ifndef CONFIG_INTERRUPT_SANITIZE_REGISTERS
>>       cmpdi    r3,0
>>       bne-    .Lrestore_nvgprs_\srr
>