From: Christoph Hellwig <hch@lst.de>
To: Hannes Reinecke <hare@suse.de>
Cc: "Martin K. Petersen" <martin.petersen@oracle.com>,
Christoph Hellwig <hch@lst.de>,
James Bottomley <james.bottomley@hansenpartnership.com>,
Sumit Saxena <sumit.saxena@broadcom.com>,
linux-scsi@vger.kernel.org, Hannes Reinecke <hare@suse.com>
Subject: Re: [PATCH 1/5] megaraid_sas: switch to pci_alloc_irq_vectors
Date: Mon, 14 Nov 2016 13:48:55 +0100 [thread overview]
Message-ID: <20161114124855.GC14653@lst.de> (raw)
In-Reply-To: <1478857492-4581-2-git-send-email-hare@suse.de>
> if (instance->msix_vectors)
> for (i = 0; i < instance->msix_vectors; i++) {
> + free_irq(pci_irq_vector(instance->pdev, i),
> &instance->irq_context[i]);
> }
> else
> + free_irq(pci_irq_vector(instance->pdev, 0),
> + &instance->irq_context[0]);
> }
Don't forget to replace the call to pci_disable_msix with one to pci_free_irq_vectors.
>
> /**
> @@ -5018,6 +5004,7 @@ static int megasas_init_fw(struct megasas_instance *instance)
> int i, loop, fw_msix_count = 0;
> struct IOV_111 *iovPtr;
> struct fusion_context *fusion;
> + int irq_flags = PCI_IRQ_MSIX;
>
> fusion = instance->ctrl_context;
>
> @@ -5134,15 +5121,18 @@ static int megasas_init_fw(struct megasas_instance *instance)
> /* Don't bother allocating more MSI-X vectors than cpus */
> instance->msix_vectors = min(instance->msix_vectors,
> (unsigned int)num_online_cpus());
> - for (i = 0; i < instance->msix_vectors; i++)
> - instance->msixentry[i].entry = i;
> - i = pci_enable_msix_range(instance->pdev, instance->msixentry,
> - 1, instance->msix_vectors);
> + if (smp_affinity_enable)
> + irq_flags |= PCI_IRQ_AFFINITY;
> + i = pci_alloc_irq_vectors(instance->pdev, 1,
> + instance->msix_vectors, irq_flags);
> if (i > 0)
> instance->msix_vectors = i;
> else
> instance->msix_vectors = 0;
> }
> + i = pci_alloc_irq_vectors(instance->pdev, 1, 1, PCI_IRQ_LEGACY);
> + if (i < 0)
> + goto fail_setup_irqs;
This looks wrong - you have to call to pci_alloc_irq_vectors right next
to each other here, so for the MSI-X case you'll still end up calling
the second one as well, which will then fail. I think the better way
to structure the driver would be to do the following here.
- Rename the msix_vectors field to irq_vectors, and make sure it's
initialized to 1 for non-MSI-X capable adapters.
- Have a single call to pci_alloc_irq_vectors here.
- Have all the request_irq/free_irq code iterate over ->irq_vectors
instead of duplicating it
- use pdev->msix_enabled for the few cases that need to check for
MSI-X mode specificly.
> /* Now re-enable MSI-X */
> - if (instance->msix_vectors &&
> - pci_enable_msix_exact(instance->pdev, instance->msixentry,
> - instance->msix_vectors))
> + if (instance->msix_vectors)
> + irq_flags = PCI_IRQ_MSIX;
> + if (smp_affinity_enable)
> + irq_flags |= PCI_IRQ_AFFINITY;
> + rval = pci_alloc_irq_vectors(instance->pdev, 1,
> + instance->msix_vectors ?
> + instance->msix_vectors : 1, irq_flags);
> + if (rval < 0)
The old code is doing a pci_enable_msix_exact so you also need to pass
the number of vectors as the first argument here.
next prev parent reply other threads:[~2016-11-14 12:48 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-11-11 9:44 [PATCH 0/5] megaraid_sas: scsi-mq support Hannes Reinecke
2016-11-11 9:44 ` [PATCH 1/5] megaraid_sas: switch to pci_alloc_irq_vectors Hannes Reinecke
2016-11-11 11:32 ` Sumit Saxena
2016-11-11 14:59 ` Hannes Reinecke
2016-11-14 12:33 ` Christoph Hellwig
2016-11-14 12:48 ` Christoph Hellwig [this message]
2016-11-11 9:44 ` [PATCH 2/5] megaraid_sas: avoid calling megasas_lookup_instance() Hannes Reinecke
2016-11-11 10:46 ` Sumit Saxena
2016-11-11 9:44 ` [PATCH 3/5] megaraid_sas: do not crash on invalid completion Hannes Reinecke
2016-11-11 11:51 ` Sumit Saxena
2016-11-11 15:07 ` Hannes Reinecke
2016-11-11 9:44 ` [PATCH 4/5] megaraid_sas: scsi-mq support Hannes Reinecke
2016-11-11 11:26 ` kbuild test robot
2016-11-11 11:56 ` Sumit Saxena
2016-11-14 11:07 ` Kashyap Desai
2016-11-11 9:44 ` [PATCH 5/5] megaraid_sas: add mmio barrier after register writes Hannes Reinecke
2016-11-11 10:47 ` Sumit Saxena
2016-11-18 15:53 ` Tomas Henzl
2016-11-18 16:48 ` Kashyap Desai
2016-11-21 15:57 ` Tomas Henzl
2016-11-30 6:14 ` Kashyap Desai
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20161114124855.GC14653@lst.de \
--to=hch@lst.de \
--cc=hare@suse.com \
--cc=hare@suse.de \
--cc=james.bottomley@hansenpartnership.com \
--cc=linux-scsi@vger.kernel.org \
--cc=martin.petersen@oracle.com \
--cc=sumit.saxena@broadcom.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).