From: Vitaly Kuznetsov <vkuznets@redhat.com>
To: Thomas Gleixner <tglx@linutronix.de>, x86@kernel.org
Cc: Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
"H. Peter Anvin" <hpa@zytor.com>,
linux-kernel@vger.kernel.org
Subject: Re: [PATCH RFC 2/2] genirq/matrix: WARN_ON_ONCE() when cm->allocated/m->total_allocated go negative
Date: Thu, 18 Mar 2021 08:58:55 +0100 [thread overview]
Message-ID: <87eegc7v0g.fsf@vitty.brq.redhat.com> (raw)
In-Reply-To: <87eegdy0qh.fsf@nanos.tec.linutronix.de>
Thomas Gleixner <tglx@linutronix.de> writes:
> On Fri, Feb 19 2021 at 12:31, Vitaly Kuznetsov wrote:
>
>> When irq_matrix_assign()/irq_matrix_free() calls get unsynced, weird
>> effects are possible, e.g. when cm->allocated goes negative CPU hotplug
>> may get blocked. Add WARN_ON_ONCE() to simplify detecting such situations.
>>
>> Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
>> ---
>> kernel/irq/matrix.c | 11 +++++++++--
>> 1 file changed, 9 insertions(+), 2 deletions(-)
>>
>> diff --git a/kernel/irq/matrix.c b/kernel/irq/matrix.c
>> index 651a4ad6d711..2438a4f9d726 100644
>> --- a/kernel/irq/matrix.c
>> +++ b/kernel/irq/matrix.c
>> @@ -189,7 +189,9 @@ void irq_matrix_assign_system(struct irq_matrix *m, unsigned int bit,
>> set_bit(bit, m->system_map);
>> if (replace) {
>> BUG_ON(!test_and_clear_bit(bit, cm->alloc_map));
>> + WARN_ON_ONCE(!cm->allocated);
>> cm->allocated--;
>> + WARN_ON_ONCE(!m->total_allocated);
>
> This hunk is not really useful. It already dies when the bit is not set
> in the alloc map.
This was to check for the hypothetical issue when then number of bits
set get out of sync with 'total_allocated' counter -- which is likely
impossible today but could maybe be useful as a future proof. In case
this seems to be too much I'm not against dropping it.
>
>> m->total_allocated--;
>> }
>> if (bit >= m->alloc_start && bit < m->alloc_end)
>> @@ -424,12 +426,17 @@ void irq_matrix_free(struct irq_matrix *m, unsigned int cpu,
>> return;
>>
>> clear_bit(bit, cm->alloc_map);
>> + WARN_ON_ONCE(!cm->allocated);
>> cm->allocated--;
>
> WARN and then decrement is not necessarily any better than just
> decrementing unconditionally. It's just more noisy.
>
> Why would you let the counter wrap into negative space if you already
> know it's 0?
>
> There is a way more useful way to handle this. In such a case the bit is
> NOT set in the alloc map. So:
>
> if (!WARN_ON_ONCE(test_and_clear_bit(bit, cm->alloc_map)))
> return;
>
> would have caught the problem at hand nicely and let the machine survive
> while just throwing warns and continuing is broken to begin with.
Thanks, I like the idea. I didn't do that probably because the problem
which triggered me to write these patches wasn't fatal, it was just
causing CPU0 offlining to fail.
>
> Thanks,
>
> tglx
>
--
Vitaly
next prev parent reply other threads:[~2021-03-18 7:59 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-02-19 11:30 [PATCH RFC 0/2] x86/apic: Avoid cm->allocated going negative in irq_matrix Vitaly Kuznetsov
2021-02-19 11:31 ` [PATCH RFC 1/2] x86/apic: Do not make an exception for PIC_CASCADE_IR when marking legacy irqs " Vitaly Kuznetsov
2021-03-17 20:14 ` Thomas Gleixner
2021-03-17 21:18 ` Thomas Gleixner
2021-03-17 21:40 ` Thomas Gleixner
2021-03-17 22:52 ` Thomas Gleixner
2021-03-18 8:47 ` Vitaly Kuznetsov
2021-03-18 8:29 ` Vitaly Kuznetsov
2021-03-18 10:10 ` Thomas Gleixner
2021-02-19 11:31 ` [PATCH RFC 2/2] genirq/matrix: WARN_ON_ONCE() when cm->allocated/m->total_allocated go negative Vitaly Kuznetsov
2021-03-17 20:38 ` Thomas Gleixner
2021-03-18 7:58 ` Vitaly Kuznetsov [this message]
2021-03-18 19:29 ` Thomas Gleixner
2021-03-19 9:59 ` Vitaly Kuznetsov
2021-03-11 10:19 ` [PATCH RFC 0/2] x86/apic: Avoid cm->allocated going negative in irq_matrix Vitaly Kuznetsov
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=87eegc7v0g.fsf@vitty.brq.redhat.com \
--to=vkuznets@redhat.com \
--cc=bp@alien8.de \
--cc=hpa@zytor.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@redhat.com \
--cc=tglx@linutronix.de \
--cc=x86@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).