From: James Morse <james.morse@arm.com>
To: Borislav Petkov <bp@alien8.de>, Tyler Baicar <tbaicar@codeaurora.org>
Cc: linux-acpi@vger.kernel.org, kvmarm@lists.cs.columbia.edu,
linux-arm-kernel@lists.infradead.org, linux-mm@kvack.org,
Marc Zyngier <marc.zyngier@arm.com>,
Christoffer Dall <cdall@kernel.org>,
Will Deacon <will.deacon@arm.com>,
Catalin Marinas <catalin.marinas@arm.com>,
Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>,
Rafael Wysocki <rjw@rjwysocki.net>, Len Brown <lenb@kernel.org>,
Tony Luck <tony.luck@intel.com>,
Dongjiu Geng <gengdongjiu@huawei.com>,
Xie XiuQi <xiexiuqi@huawei.com>,
Punit Agrawal <punit.agrawal@arm.com>,
jonathan.zhang@cavium.com, Thomas Gleixner <tglx@linutronix.de>
Subject: Re: [PATCH v3 07/12] ACPI / APEI: Make the nmi_fixmap_idx per-ghes to allow multiple in_nmi() users
Date: Wed, 16 May 2018 15:51:14 +0100 [thread overview]
Message-ID: <7c871e15-689c-226d-760d-dd92614de2e9@arm.com> (raw)
In-Reply-To: <20180516110348.GA17092@pd.tnic>
Hi Borislav,
On 16/05/18 12:05, Borislav Petkov wrote:
> On Tue, May 08, 2018 at 09:45:01AM +0100, James Morse wrote:
>> NOTIFY_NMI is x86's NMI, arm doesn't have anything that behaves in the same way,
>> so doesn't use it. The equivalent notifications with NMI-like behaviour are:
>> * SEA (synchronous external abort)
>> * SEI (SError Interrupt)
>> * SDEI (software delegated exception interface)
> > Oh wow, three! :)
The first two overload existing architectural behavior, the third improves all
this with a third standard option. Its the standard story!
>> Alternatively, I can put the fixmap-page and spinlock in some 'struct
>> ghes_notification' that only the NMI-like struct-ghes need. This is just moving
>> the indirection up a level, but it does pair the lock with the thing it locks,
>> and gets rid of assigning spinlock pointers.
>
> Keeping the lock and what it protects in one place certainly sounds
> better.
Yup, I was about to post a v4...
> I guess you could so something like this:
>
> struct ghes_fixmap {
> union {
> raw_spinlock_t nmi_lock;
> spinlock_t lock;
> };
(heh, spinlock_t already contains a raw_spinlock_t)
> void __iomem *(map)(struct ghes_fixmap *);
> };
>
> and assign the proper ghes_ioremap function to ->map.
The function pointer is a problem because SDEI is effectively two notification
methods. Critical can interrupt normal. I'd really like to keep the differences
buried in the SDEI driver.
v4 has a separate structure for the fixmap-entry and lock, which
ghes_copy_tofrom_phys() reaches into if in_nmi().
> The spin_lock_irqsave() call in ghes_copy_tofrom_phys() is kinda
> questionable. Because we should have disabled interrupts so that you can
> do
>
> spin_lock(map->lock);
I thought this was for the polled driver, but that must be backed by an
interrupt too...
linux/timer.h has:
| * An irqsafe timer is executed with IRQ disabled and it's safe to wait for
| * the completion of the running instance from IRQ handlers, for example,
| * by calling del_timer_sync().
| *
| * Note: The irq disabled callback execution is a special case for
| * workqueue locking issues. It's not meant for executing random crap
| * with interrupts disabled. Abuse is monitored!
This irq-disable behaviour is controlled by the flags field:
| #define TIMER_DEFERRABLE 0x00080000
| #define TIMER_IRQSAFE 0x00200000
and ghes_probe() does this:
| timer_setup(&ghes->timer, ghes_poll_func, TIMER_DEFERRABLE);
So I think the ghes_poll_func() can be called with IRQs unmasked, hence the
spin_lock_irqsave().
> Except that we do get called with IRQs on and looking at that call of
> ghes_proc() at the end of ghes_probe(), that's a deadlock waiting to
> happen.
>
> And that comes from:
>
> 77b246b32b2c ("acpi: apei: check for pending errors when probing GHES entries")
>
> Tyler, this can't work in any context: imagine the GHES NMI or IRQ or
> the timer fires while that ghes_proc() runs...
I thought this was safe because its just ghes_copy_tofrom_phys()s access to the
fixmap slots that needs mutual exclusion.
Polled and all the IRQ flavours are kept apart by the spin_lock_irqsave(), and
the NMIs have their own fixmap entry. (This is fine until there is more than
once source of NMI)
Thanks,
James
next prev parent reply other threads:[~2018-05-16 14:54 UTC|newest]
Thread overview: 26+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-04-27 15:34 [PATCH v3 00/12] APEI in_nmi() rework and arm64 SDEI wire-up James Morse
2018-04-27 15:34 ` [PATCH v3 01/12] ACPI / APEI: Move the estatus queue code up, and under its own ifdef James Morse
2018-05-01 10:43 ` Punit Agrawal
2018-05-01 12:50 ` James Morse
2018-05-05 9:58 ` Borislav Petkov
2018-04-27 15:35 ` [PATCH v3 02/12] ACPI / APEI: Generalise the estatus queue's add/remove and notify code James Morse
2018-05-05 10:12 ` Borislav Petkov
2018-04-27 15:35 ` [PATCH v3 03/12] ACPI / APEI: don't wait to serialise with oops messages when panic()ing James Morse
2018-04-27 15:35 ` [PATCH v3 04/12] ACPI / APEI: Switch NOTIFY_SEA to use the estatus queue James Morse
2018-04-27 15:35 ` [PATCH v3 05/12] KVM: arm/arm64: Add kvm_ras.h to collect kvm specific RAS plumbing James Morse
2018-04-27 15:35 ` [PATCH v3 06/12] arm64: KVM/mm: Move SEA handling behind a single 'claim' interface James Morse
2018-04-27 15:35 ` [PATCH v3 07/12] ACPI / APEI: Make the nmi_fixmap_idx per-ghes to allow multiple in_nmi() users James Morse
2018-05-05 12:27 ` Borislav Petkov
2018-05-08 8:45 ` James Morse
2018-05-16 11:05 ` Borislav Petkov
2018-05-16 14:51 ` James Morse [this message]
2018-05-17 13:36 ` Borislav Petkov
2018-05-17 18:11 ` James Morse
2018-05-16 15:38 ` Tyler Baicar
2018-05-17 13:39 ` Borislav Petkov
2018-04-27 15:35 ` [PATCH v3 08/12] ACPI / APEI: Split fixmap pages for arm64 NMI-like notifications James Morse
2018-04-27 15:35 ` [PATCH v3 09/12] firmware: arm_sdei: Add ACPI GHES registration helper James Morse
2018-04-27 15:35 ` [PATCH v3 10/12] ACPI / APEI: Add support for the SDEI GHES Notification type James Morse
2018-04-27 15:35 ` [PATCH v3 11/12] mm/memory-failure: increase queued recovery work's priority James Morse
2018-04-27 15:35 ` [PATCH v3 12/12] arm64: acpi: Make apei_claim_sea() synchronise with APEI's irq work James Morse
2018-05-01 20:15 ` [PATCH v3 00/12] APEI in_nmi() rework and arm64 SDEI wire-up Tyler Baicar
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=7c871e15-689c-226d-760d-dd92614de2e9@arm.com \
--to=james.morse@arm.com \
--cc=bp@alien8.de \
--cc=catalin.marinas@arm.com \
--cc=cdall@kernel.org \
--cc=gengdongjiu@huawei.com \
--cc=jonathan.zhang@cavium.com \
--cc=kvmarm@lists.cs.columbia.edu \
--cc=lenb@kernel.org \
--cc=linux-acpi@vger.kernel.org \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-mm@kvack.org \
--cc=marc.zyngier@arm.com \
--cc=n-horiguchi@ah.jp.nec.com \
--cc=punit.agrawal@arm.com \
--cc=rjw@rjwysocki.net \
--cc=tbaicar@codeaurora.org \
--cc=tglx@linutronix.de \
--cc=tony.luck@intel.com \
--cc=will.deacon@arm.com \
--cc=xiexiuqi@huawei.com \
/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).