linux-kernel - Re: [PATCH v4 2/2] PCI/aer_inject: Convert inject_lock to raw_spinlock

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 for Android: free password hash cracker in your pocket

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

Message-ID: <aQydxtOtgPTPxL_9@2a2a0ba7cec8>
Date: Thu, 6 Nov 2025 13:08:22 +0000
From: Guangbo Cui <jckeep.cuiguangbo@...il.com>
To: Peter Zijlstra <peterz@...radead.org>
Cc: Sebastian Andrzej Siewior <bigeasy@...utronix.de>,
	Clark Williams <clrkwllms@...nel.org>,
	Steven Rostedt <rostedt@...dmis.org>,
	Ingo Molnar <mingo@...hat.com>, Will Deacon <will@...nel.org>,
	Boqun Feng <boqun.feng@...il.com>,
	Thomas Gleixner <tglx@...utronix.de>,
	Bjorn Helgaas <bhelgaas@...gle.com>,
	Waiman Long <longman@...hat.com>, linux-rt-devel@...ts.linux.dev,
	linux-kernel@...r.kernel.org, linux-pci@...r.kernel.org
Subject: Re: [PATCH v4 2/2] PCI/aer_inject: Convert inject_lock to
 raw_spinlock_t

On Mon, Nov 03, 2025 at 08:27:09PM +0100, Peter Zijlstra wrote:
> On Mon, Nov 03, 2025 at 08:21:20PM +0100, Peter Zijlstra wrote:
> > On Sun, Nov 02, 2025 at 10:57:06AM +0000, Guangbo Cui wrote:
> > > The AER injection path may run in interrupt-disabled context while
> > > holding inject_lock. On PREEMPT_RT kernels, spinlock_t becomes a
> > > sleeping lock, so it must not be taken while a raw_spinlock_t is held.
> > > Doing so violates lock ordering rules and trigger lockdep reports
> > > such as “Invalid wait context”.
> > > 
> > > Convert inject_lock to raw_spinlock_t to ensure non-sleeping locking
> > > semantics. The protected list is bounded and used only for debugging,
> > > so raw locking will not cause latency issues.
> > 
> > Bounded how?
> > 
> +	scoped_guard (raw_spinlock_irqsave, &inject_lock) {
> +		if (ops == &aer_inj_pci_ops)
> +			break;
> +		pci_bus_ops_init(bus_ops, bus, ops);
> +		list_add(&bus_ops->list, &pci_bus_ops_list);
> +		bus_ops = NULL;
> +	}

I found that there are two styles of calling scoped_guard in the kernel:

  1. scoped_guard (...)

  2. scoped_guard(...)

Is there any coding convention or guideline regarding this?

> +				rperr->root_status |= PCI_ERR_ROOT_COR_RCV;
> +			rperr->source_id &= 0xffff0000;
> +			rperr->source_id |= PCI_DEVID(einj->bus, devfn);
> +		}
> +		if (einj->uncor_status) {
> +			if (rperr->root_status & PCI_ERR_ROOT_UNCOR_RCV)
> +				rperr->root_status |= PCI_ERR_ROOT_MULTI_UNCOR_RCV;
> +			if (sever & einj->uncor_status) {
> +				rperr->root_status |= PCI_ERR_ROOT_FATAL_RCV;
> +				if (!(rperr->root_status & PCI_ERR_ROOT_UNCOR_RCV))
> +					rperr->root_status |= PCI_ERR_ROOT_FIRST_FATAL;
> +			} else
> +				rperr->root_status |= PCI_ERR_ROOT_NONFATAL_RCV;
> +			rperr->root_status |= PCI_ERR_ROOT_UNCOR_RCV;
> +			rperr->source_id &= 0x0000ffff;
> +			rperr->source_id |= PCI_DEVID(einj->bus, devfn) << 16;
> +		}
>  	}
> -	raw_spin_unlock_irqrestore(&inject_lock, flags);
>  
>  	if (aer_mask_override) {
>  		pci_write_config_dword(dev, pos_cap_err + PCI_ERR_COR_MASK,

LGTM, If there are no objections, I’ll include these two patches in the
next version of the patchset and add your Signed-off-by tag.


Best regards,
Guangbo