[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <8a9bad577d4e4b019cb22e7e59f742b2@intel.com>
Date: Sat, 18 Apr 2020 07:20:02 +0000
From: "Winkler, Tomas" <tomas.winkler@...el.com>
To: Benjamin Lee <ben@...1l1.com>,
"Usyskin, Alexander" <alexander.usyskin@...el.com>,
Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: RE: [PATCH] mei: me: fix irq number stored in hw struct
> Commit 261b3e1f2a01 ("mei: me: store irq number in the hw struct.") stores
> the irq number in the hw struct before MSI is enabled. This caused a regression
> for mei_me_synchronize_irq() waiting for the wrong irq number. On my laptop
> this causes a hang on shutdown. Fix the issue by storing the irq number after
> enabling MSI.
Thanks a lot for cracking this down, this should go to stable (5.5) as well.
We should be probably calling hw->irq = pci_irq_vector(pdev, 0); as pci_enable_msi was deprecated.
Will try to address this in a follow up patch.
> Fixes: 261b3e1f2a01 ("mei: me: store irq number in the hw struct.")
> Signed-off-by: Benjamin Lee <ben@...1l1.com>
> ---
> drivers/misc/mei/pci-me.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/misc/mei/pci-me.c b/drivers/misc/mei/pci-me.c index
> 3d21c38e2dbb..0c390fe421ad 100644
> --- a/drivers/misc/mei/pci-me.c
> +++ b/drivers/misc/mei/pci-me.c
> @@ -203,11 +203,12 @@ static int mei_me_probe(struct pci_dev *pdev, const
> struct pci_device_id *ent)
> }
> hw = to_me_hw(dev);
> hw->mem_addr = pcim_iomap_table(pdev)[0];
> - hw->irq = pdev->irq;
> hw->read_fws = mei_me_read_fws;
>
> pci_enable_msi(pdev);
>
> + hw->irq = pdev->irq;
> +
> /* request and enable interrupt */
> irqflags = pci_dev_msi_enabled(pdev) ? IRQF_ONESHOT :
> IRQF_SHARED;
>
> --
> 2.26.1
Powered by blists - more mailing lists