linux-kernel - Re: [patch V2 05/10] PCI/MSI: Switch to MSI descriptor locking to guard()

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: <20250314094237.GX5880@noisy.programming.kicks-ass.net>
Date: Fri, 14 Mar 2025 10:42:37 +0100
From: Peter Zijlstra <peterz@...radead.org>
To: Thomas Gleixner <tglx@...utronix.de>
Cc: Jonathan Cameron <Jonathan.Cameron@...wei.com>,
	LKML <linux-kernel@...r.kernel.org>, Marc Zyngier <maz@...nel.org>,
	Bjorn Helgaas <bhelgaas@...gle.com>, linux-pci@...r.kernel.org,
	Nishanth Menon <nm@...com>, Dhruva Gole <d-gole@...com>,
	Tero Kristo <kristo@...nel.org>,
	Santosh Shilimkar <ssantosh@...nel.org>,
	Logan Gunthorpe <logang@...tatee.com>,
	Dave Jiang <dave.jiang@...el.com>, Jon Mason <jdmason@...zu.us>,
	Allen Hubbe <allenbh@...il.com>, ntb@...ts.linux.dev,
	Michael Kelley <mhklinux@...look.com>, Wei Liu <wei.liu@...nel.org>,
	Haiyang Zhang <haiyangz@...rosoft.com>,
	linux-hyperv@...r.kernel.org, Wei Huang <wei.huang2@....com>,
	Manivannan Sadhasivam <manivannan.sadhasivam@...aro.org>,
	"James E.J. Bottomley" <James.Bottomley@...senpartnership.com>,
	"Martin K. Petersen" <martin.petersen@...cle.com>,
	linux-scsi@...r.kernel.org,
	Jonathan Cameron <Jonathan.Cameron@...ei.com>
Subject: Re: [patch V2 05/10] PCI/MSI: Switch to MSI descriptor locking to
 guard()

On Thu, Mar 13, 2025 at 06:55:12PM +0100, Thomas Gleixner wrote:
> On Thu, Mar 13 2025 at 15:50, Jonathan Cameron wrote:
> >> +	guard(msi_descs_lock)(&dev->dev);
> >> +	int ret = __msix_setup_interrupts(dev, entries, nvec, masks);
> >> +	if (ret)
> >> +		pci_free_msi_irqs(dev);
> >
> > It's not immediately obvious what this is undoing (i.e. where the alloc
> > is).  I think it is at least mostly the pci_msi_setup_msi_irqs in
> > __msix_setup_interrupts
> 
> It's a universal cleanup for all possible error cases.
> 
> > Why not handle the error in __msix_setup_interrupts and make that function
> > side effect free.  Does require gotos but in a function that isn't
> > doing any cleanup magic so should be fine.
> 
> I had the gotos first and then hated them. But you are right, it's better
> to have them than having the magic clean up at the call site.
> 
> I'll fold the delta patch below.
> 
> Thanks,
> 
>         tglx
> ---
> --- a/drivers/pci/msi/msi.c
> +++ b/drivers/pci/msi/msi.c
> @@ -671,19 +671,23 @@ static int __msix_setup_interrupts(struc
>  	int ret = msix_setup_msi_descs(dev, entries, nvec, masks);
>  
>  	if (ret)
> -		return ret;
> +		goto fail;
>  
>  	ret = pci_msi_setup_msi_irqs(dev, nvec, PCI_CAP_ID_MSIX);
>  	if (ret)
> -		return ret;
> +		goto fail;
>  
>  	/* Check if all MSI entries honor device restrictions */
>  	ret = msi_verify_entries(dev);
>  	if (ret)
> -		return ret;
> +		goto fail;
>  
>  	msix_update_entries(dev, entries);
>  	return 0;
> +
> +fail:
> +	pci_free_msi_irqs(dev);
> +	return ret;
>  }

How about something like:

int __msix_setup_interrupts(struct device *_dev,... )
{
	struct device *dev __free(msi_irqs) = _dev;

	int ret = msix_setup_msi_descs(dev, entries, nvec, masks);
	if (ret)
		return ret;

	ret = pci_msi_setup_msi_irqs(dev, nvec, PCI_CAP_ID_MSIX);
	if (ret)
		return ret;

	/* Check if all MSI entries honor device restrictions */
	ret = msi_verify_entries(dev);
	if (ret)
		return ret;

	retain_ptr(dev);
	msix_update_entries(dev, entries);
	return 0;
}

?