[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CA+sq2CdsGPNwU0kKQUtRgxKvGTDoYE0yLVC+fSgc3DGbjOjWVg@mail.gmail.com>
Date: Mon, 27 Mar 2017 21:33:32 +0530
From: Sunil Kovvuri <sunil.kovvuri@...il.com>
To: Christoph Hellwig <hch@....de>
Cc: Bjorn Helgaas <bhelgaas@...gle.com>,
"David S. Miller" <davem@...emloft.net>, netanel@...apurnalabs.com,
jcliburn@...il.com, chris.snook@...il.com,
Sunil Goutham <sgoutham@...ium.com>,
Robert Richter <rric@...nel.org>,
Linux Netdev List <netdev@...r.kernel.org>,
linux-pci <linux-pci@...r.kernel.org>
Subject: Re: [PATCH 1/5] net: thunderx: switch to pci_alloc_irq_vectors
On Mon, Mar 27, 2017 at 1:59 PM, Christoph Hellwig <hch@....de> wrote:
> Remove the deprecated pci_enable_msix API in favour of it's successor.
>
> Signed-off-by: Christoph Hellwig <hch@....de>
> ---
> drivers/net/ethernet/cavium/thunder/nic_main.c | 75 ++++++--------------------
> 1 file changed, 15 insertions(+), 60 deletions(-)
> - nic->num_vec = pci_msix_vec_count(nic->pdev);
> -
> - ret = pci_enable_msix(nic->pdev, nic->msix_entries, nic->num_vec);
> - if (ret) {
> + /* Enable MSI-X */
> + ret = pci_alloc_irq_vectors(nic->pdev, nic->num_vec, nic->num_vec,
> + PCI_IRQ_MSIX);
nic->num_vec will always be zero here.
> + while (--i >= 0)
> + free_irq(pci_irq_vector(nic->pdev, i), nic);
> + pci_free_irq_vectors(nic->pdev);
> return ret;
IRQ handlers are not registered for all vectors to free them unconditionally
> + int i;
> +
> + for (i = 0; i < nic->num_vec; i++)
> + free_irq(pci_irq_vector(nic->pdev, i), nic);
> + pci_free_irq_vectors(nic->pdev);
> }
Same as here.
How urgent is this ?
Can we comeup with a proper patch to fix this in couple of weeks, instead of
going with these untested patches ?
Powered by blists - more mailing lists