linux-kernel - Re: BUG: INTx is assered unexpectly when unload AHCI driver with MSIx support.

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>] [day] [month] [year] [list]

Message-ID: <20160706135236.GG3262@mtj.duckdns.org>
Date:	Wed, 6 Jul 2016 09:52:36 -0400
From:	"tj@...nel.org" <tj@...nel.org>
To:	Pang Raymond <Raymond_rule@...mail.com>
Cc:	linux ide <linux-ide@...r.kernel.org>,
	linux kernel <linux-kernel@...r.kernel.org>
Subject: Re: BUG: INTx is assered unexpectly when unload AHCI driver with
 MSIx support.

Hello,

(Can you please flow the text a bit below 80 column when you reply?)

On Wed, Jul 06, 2016 at 02:51:47AM +0000, Pang Raymond wrote:
> Why does the phenomenon appear?
> 	Two factors cause this problem.
> 		1. In MSIx's irq handler
> 		   (ahci_multi_irqs_intr_hard()), we do not clear
> 		   GHC.IS register which presents which ports need
> 		   interrupt service.  static irqreturn_t
> 		   ahci_multi_irqs_intr_hard(int irq, void
> 		   *dev_instance)
> 			{
> 				//  omitting unconcerned codes here
> 				// ...				
> 				// here we just clear PxIS register, not clear GHC.IS
> 				status = readl(port_mmio + PORT_IRQ_STAT);
> 				writel(status, port_mmio + PORT_IRQ_STAT);
> 				// ...
> 			}
> 		2. In PCI subsystem, after Disable MSIx, INTx is
>		   enabled automatically.
> 			void pci_msix_shutdown(struct pci_dev *dev)
> 			{
> 				// omitting unconcerned codes here
> 				// ...				
> 				// after Disable MSIx, enable INTx imediately
> 				pci_msix_clear_and_set_ctrl(dev, PCI_MSIX_FLAGS_ENABLE, 0);
> 				pci_intx_for_msi(dev, 1);
> 				dev->msix_enabled = 0;
> 				pcibios_alloc_irq(dev);
> 			}
>
> When Kernel enables INTx(I mean clearing PCI configure space
> Rx04.bit10), controller will find GHC.IS is not Zero (because driver
> haven't cleared GHC.IS after servicing MSIx interrupt), so it
> believes some SATA ports need interrupt servicing and asserts INTx
> interrupt.

Looks like you debugged the problem.  Care to write up and test a
patch?

Thanks a lot!

-- 
tejun