* [PATCH] ntb: Use pci_enable_msix_range() instead of pci_enable_msix() @ 2014-02-19 10:15 Alexander Gordeev 2014-02-19 18:09 ` Jon Mason 0 siblings, 1 reply; 5+ messages in thread From: Alexander Gordeev @ 2014-02-19 10:15 UTC (permalink / raw) To: linux-kernel; +Cc: Alexander Gordeev, Jon Mason, linux-pci As result of deprecation of MSI-X/MSI enablement functions pci_enable_msix() and pci_enable_msi_block() all drivers using these two interfaces need to be updated to use the new pci_enable_msi_range() and pci_enable_msix_range() interfaces. Signed-off-by: Alexander Gordeev <agordeev@redhat.com> Cc: Jon Mason <jon.mason@intel.com> Cc: linux-pci@vger.kernel.org --- drivers/ntb/ntb_hw.c | 43 ++++++++++++++++--------------------------- drivers/ntb/ntb_hw.h | 2 -- 2 files changed, 16 insertions(+), 29 deletions(-) diff --git a/drivers/ntb/ntb_hw.c b/drivers/ntb/ntb_hw.c index 170e8e6..fda37eb 100644 --- a/drivers/ntb/ntb_hw.c +++ b/drivers/ntb/ntb_hw.c @@ -1085,21 +1085,15 @@ static int ntb_setup_msix(struct ntb_device *ndev) struct msix_entry *msix; int msix_entries; int rc, i; - u16 val; - if (!pdev->msix_cap) { - rc = -EIO; - goto err; - } - - rc = pci_read_config_word(pdev, pdev->msix_cap + PCI_MSIX_FLAGS, &val); - if (rc) + rc = pci_msix_vec_count(pdev); + if (rc < 0) { goto err; - - msix_entries = msix_table_size(val); - if (msix_entries > ndev->limits.msix_cnt) { + } else if (rc > ndev->limits.msix_cnt) { rc = -EINVAL; goto err; + } else { + msix_entries = rc; } ndev->msix_entries = kmalloc(sizeof(struct msix_entry) * msix_entries, @@ -1112,26 +1106,21 @@ static int ntb_setup_msix(struct ntb_device *ndev) for (i = 0; i < msix_entries; i++) ndev->msix_entries[i].entry = i; - rc = pci_enable_msix(pdev, ndev->msix_entries, msix_entries); - if (rc < 0) - goto err1; - if (rc > 0) { + if (ndev->hw_type != BWD_HW) /* On SNB, the link interrupt is always tied to 4th vector. If * we can't get all 4, then we can't use MSI-X. */ - if (ndev->hw_type != BWD_HW) { - rc = -EIO; - goto err1; - } - - dev_warn(&pdev->dev, - "Only %d MSI-X vectors. Limiting the number of queues to that number.\n", - rc); + rc = pci_enable_msix_range(pdev, ndev->msix_entries, + msix_entries, msix_entries); + else + rc = pci_enable_msix_range(pdev, ndev->msix_entries, + 1, msix_entries); + if (rc < 0) + goto err1; + else if (rc < msix_entries) { + dev_warn(&pdev->dev, "Only %d MSI-X vectors. " + "Limiting the number of queues to that number.\n", rc); msix_entries = rc; - - rc = pci_enable_msix(pdev, ndev->msix_entries, msix_entries); - if (rc) - goto err1; } for (i = 0; i < msix_entries; i++) { diff --git a/drivers/ntb/ntb_hw.h b/drivers/ntb/ntb_hw.h index bbdb7ed..d307107 100644 --- a/drivers/ntb/ntb_hw.h +++ b/drivers/ntb/ntb_hw.h @@ -60,8 +60,6 @@ #define PCI_DEVICE_ID_INTEL_NTB_SS_HSX 0x2F0F #define PCI_DEVICE_ID_INTEL_NTB_B2B_BWD 0x0C4E -#define msix_table_size(control) ((control & PCI_MSIX_FLAGS_QSIZE)+1) - #ifndef readq static inline u64 readq(void __iomem *addr) { -- 1.7.7.6 ^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH] ntb: Use pci_enable_msix_range() instead of pci_enable_msix() 2014-02-19 10:15 [PATCH] ntb: Use pci_enable_msix_range() instead of pci_enable_msix() Alexander Gordeev @ 2014-02-19 18:09 ` Jon Mason 2014-02-19 19:22 ` Alexander Gordeev 2014-02-20 13:22 ` Alexander Gordeev 0 siblings, 2 replies; 5+ messages in thread From: Jon Mason @ 2014-02-19 18:09 UTC (permalink / raw) To: Alexander Gordeev; +Cc: linux-kernel, linux-pci On Wed, Feb 19, 2014 at 11:15:21AM +0100, Alexander Gordeev wrote: > As result of deprecation of MSI-X/MSI enablement functions > pci_enable_msix() and pci_enable_msi_block() all drivers > using these two interfaces need to be updated to use the > new pci_enable_msi_range() and pci_enable_msix_range() > interfaces. > > Signed-off-by: Alexander Gordeev <agordeev@redhat.com> > Cc: Jon Mason <jon.mason@intel.com> > Cc: linux-pci@vger.kernel.org > --- > drivers/ntb/ntb_hw.c | 43 ++++++++++++++++--------------------------- > drivers/ntb/ntb_hw.h | 2 -- > 2 files changed, 16 insertions(+), 29 deletions(-) > > diff --git a/drivers/ntb/ntb_hw.c b/drivers/ntb/ntb_hw.c > index 170e8e6..fda37eb 100644 > --- a/drivers/ntb/ntb_hw.c > +++ b/drivers/ntb/ntb_hw.c > @@ -1085,21 +1085,15 @@ static int ntb_setup_msix(struct ntb_device *ndev) > struct msix_entry *msix; > int msix_entries; > int rc, i; > - u16 val; > > - if (!pdev->msix_cap) { > - rc = -EIO; > - goto err; > - } > - > - rc = pci_read_config_word(pdev, pdev->msix_cap + PCI_MSIX_FLAGS, &val); > - if (rc) > + rc = pci_msix_vec_count(pdev); Nit, you should probably use msix_entries instead of rc in this case. > + if (rc < 0) { > goto err; > - > - msix_entries = msix_table_size(val); > - if (msix_entries > ndev->limits.msix_cnt) { > + } else if (rc > ndev->limits.msix_cnt) { > rc = -EINVAL; > goto err; > + } else { > + msix_entries = rc; > } Style Nit. Braces with only one line. If you make the change above, then it is moot. > > ndev->msix_entries = kmalloc(sizeof(struct msix_entry) * msix_entries, > @@ -1112,26 +1106,21 @@ static int ntb_setup_msix(struct ntb_device *ndev) > for (i = 0; i < msix_entries; i++) > ndev->msix_entries[i].entry = i; > > - rc = pci_enable_msix(pdev, ndev->msix_entries, msix_entries); > - if (rc < 0) > - goto err1; > - if (rc > 0) { > + if (ndev->hw_type != BWD_HW) > /* On SNB, the link interrupt is always tied to 4th vector. If > * we can't get all 4, then we can't use MSI-X. > */ > - if (ndev->hw_type != BWD_HW) { > - rc = -EIO; > - goto err1; > - } > - > - dev_warn(&pdev->dev, > - "Only %d MSI-X vectors. Limiting the number of queues to that number.\n", > - rc); > + rc = pci_enable_msix_range(pdev, ndev->msix_entries, > + msix_entries, msix_entries); > + else > + rc = pci_enable_msix_range(pdev, ndev->msix_entries, > + 1, msix_entries); Actually, this must be 2 for the min. One for the Data and one for the Link. If you look a few lines after this in the original code, there is a grabbing of the last vector for the link. I realize there is currently no check for this in the driver and a potential error case occurs. I can make a separate patch to correct this issue if this patch is not going through my tree. Thanks, Jon > + if (rc < 0) > + goto err1; > + else if (rc < msix_entries) { > + dev_warn(&pdev->dev, "Only %d MSI-X vectors. " > + "Limiting the number of queues to that number.\n", rc); > msix_entries = rc; > - > - rc = pci_enable_msix(pdev, ndev->msix_entries, msix_entries); > - if (rc) > - goto err1; > } > > for (i = 0; i < msix_entries; i++) { > diff --git a/drivers/ntb/ntb_hw.h b/drivers/ntb/ntb_hw.h > index bbdb7ed..d307107 100644 > --- a/drivers/ntb/ntb_hw.h > +++ b/drivers/ntb/ntb_hw.h > @@ -60,8 +60,6 @@ > #define PCI_DEVICE_ID_INTEL_NTB_SS_HSX 0x2F0F > #define PCI_DEVICE_ID_INTEL_NTB_B2B_BWD 0x0C4E > > -#define msix_table_size(control) ((control & PCI_MSIX_FLAGS_QSIZE)+1) > - > #ifndef readq > static inline u64 readq(void __iomem *addr) > { > -- > 1.7.7.6 > ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] ntb: Use pci_enable_msix_range() instead of pci_enable_msix() 2014-02-19 18:09 ` Jon Mason @ 2014-02-19 19:22 ` Alexander Gordeev 2014-02-19 23:21 ` Jon Mason 2014-02-20 13:22 ` Alexander Gordeev 1 sibling, 1 reply; 5+ messages in thread From: Alexander Gordeev @ 2014-02-19 19:22 UTC (permalink / raw) To: Jon Mason; +Cc: linux-kernel, linux-pci On Wed, Feb 19, 2014 at 11:09:27AM -0700, Jon Mason wrote: > > diff --git a/drivers/ntb/ntb_hw.c b/drivers/ntb/ntb_hw.c > > index 170e8e6..fda37eb 100644 > > --- a/drivers/ntb/ntb_hw.c > > +++ b/drivers/ntb/ntb_hw.c > > @@ -1085,21 +1085,15 @@ static int ntb_setup_msix(struct ntb_device *ndev) > > struct msix_entry *msix; > > int msix_entries; > > int rc, i; > > - u16 val; > > > > - if (!pdev->msix_cap) { > > - rc = -EIO; > > - goto err; > > - } > > - > > - rc = pci_read_config_word(pdev, pdev->msix_cap + PCI_MSIX_FLAGS, &val); > > - if (rc) > > + rc = pci_msix_vec_count(pdev); > > Nit, you should probably use msix_entries instead of rc in this case. This way it resembles the code few lines below, so the style looks more consistent. Whichever you wish, though ;) > > + if (rc < 0) { > > goto err; > > - > > - msix_entries = msix_table_size(val); > > - if (msix_entries > ndev->limits.msix_cnt) { > > + } else if (rc > ndev->limits.msix_cnt) { > > rc = -EINVAL; > > goto err; > > + } else { > > + msix_entries = rc; > > } > > Style Nit. Braces with only one line. If you make the change above, > then it is moot. Not on the top of my head, but do we mix bracing this way? > > > > ndev->msix_entries = kmalloc(sizeof(struct msix_entry) * msix_entries, > > @@ -1112,26 +1106,21 @@ static int ntb_setup_msix(struct ntb_device *ndev) > > for (i = 0; i < msix_entries; i++) > > ndev->msix_entries[i].entry = i; > > > > - rc = pci_enable_msix(pdev, ndev->msix_entries, msix_entries); > > - if (rc < 0) > > - goto err1; > > - if (rc > 0) { > > + if (ndev->hw_type != BWD_HW) > > /* On SNB, the link interrupt is always tied to 4th vector. If > > * we can't get all 4, then we can't use MSI-X. > > */ > > - if (ndev->hw_type != BWD_HW) { > > - rc = -EIO; > > - goto err1; > > - } > > - > > - dev_warn(&pdev->dev, > > - "Only %d MSI-X vectors. Limiting the number of queues to that number.\n", > > - rc); > > + rc = pci_enable_msix_range(pdev, ndev->msix_entries, > > + msix_entries, msix_entries); > > + else > > + rc = pci_enable_msix_range(pdev, ndev->msix_entries, > > + 1, msix_entries); > > Actually, this must be 2 for the min. One for the Data and one for > the Link. If you look a few lines after this in the original code, > there is a grabbing of the last vector for the link. I realize there > is currently no check for this in the driver and a potential error > case occurs. I can make a separate patch to correct this issue if > this patch is not going through my tree. Well, I can re-post with a separate fix with this condition updated... } else if (rc < 2 || rc > ndev->limits.msix_cnt) { rc = -EINVAL; goto err; } ...and the original patch with call to pci_enable_msix_range( ..., 2, ndev->limits.msix_cnt) Will it work with both BWD_HW and non-BWD_HW hardware? > Thanks, > Jon > > > > + if (rc < 0) > > + goto err1; > > + else if (rc < msix_entries) { > > + dev_warn(&pdev->dev, "Only %d MSI-X vectors. " > > + "Limiting the number of queues to that number.\n", rc); > > msix_entries = rc; > > - > > - rc = pci_enable_msix(pdev, ndev->msix_entries, msix_entries); > > - if (rc) > > - goto err1; > > } > > > > for (i = 0; i < msix_entries; i++) { > > diff --git a/drivers/ntb/ntb_hw.h b/drivers/ntb/ntb_hw.h > > index bbdb7ed..d307107 100644 > > --- a/drivers/ntb/ntb_hw.h > > +++ b/drivers/ntb/ntb_hw.h > > @@ -60,8 +60,6 @@ > > #define PCI_DEVICE_ID_INTEL_NTB_SS_HSX 0x2F0F > > #define PCI_DEVICE_ID_INTEL_NTB_B2B_BWD 0x0C4E > > > > -#define msix_table_size(control) ((control & PCI_MSIX_FLAGS_QSIZE)+1) > > - > > #ifndef readq > > static inline u64 readq(void __iomem *addr) > > { > > -- > > 1.7.7.6 > > -- Regards, Alexander Gordeev agordeev@redhat.com ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] ntb: Use pci_enable_msix_range() instead of pci_enable_msix() 2014-02-19 19:22 ` Alexander Gordeev @ 2014-02-19 23:21 ` Jon Mason 0 siblings, 0 replies; 5+ messages in thread From: Jon Mason @ 2014-02-19 23:21 UTC (permalink / raw) To: Alexander Gordeev; +Cc: linux-kernel, linux-pci On Wed, Feb 19, 2014 at 08:22:55PM +0100, Alexander Gordeev wrote: > On Wed, Feb 19, 2014 at 11:09:27AM -0700, Jon Mason wrote: > > > diff --git a/drivers/ntb/ntb_hw.c b/drivers/ntb/ntb_hw.c > > > index 170e8e6..fda37eb 100644 > > > --- a/drivers/ntb/ntb_hw.c > > > +++ b/drivers/ntb/ntb_hw.c > > > @@ -1085,21 +1085,15 @@ static int ntb_setup_msix(struct ntb_device *ndev) > > > struct msix_entry *msix; > > > int msix_entries; > > > int rc, i; > > > - u16 val; > > > > > > - if (!pdev->msix_cap) { > > > - rc = -EIO; > > > - goto err; > > > - } > > > - > > > - rc = pci_read_config_word(pdev, pdev->msix_cap + PCI_MSIX_FLAGS, &val); > > > - if (rc) > > > + rc = pci_msix_vec_count(pdev); > > > > Nit, you should probably use msix_entries instead of rc in this case. > > This way it resembles the code few lines below, so the style looks more > consistent. Whichever you wish, though ;) It'll save the else statement below, and makes the code a little more obvious...but its not a big deal. > > > + if (rc < 0) { > > > goto err; > > > - > > > - msix_entries = msix_table_size(val); > > > - if (msix_entries > ndev->limits.msix_cnt) { > > > + } else if (rc > ndev->limits.msix_cnt) { > > > rc = -EINVAL; > > > goto err; > > > + } else { > > > + msix_entries = rc; > > > } > > > > Style Nit. Braces with only one line. If you make the change above, > > then it is moot. > > Not on the top of my head, but do we mix bracing this way? It was the preferred style once-upon-a-time to not have braces for single line if/else. I think it is really up to the driver/driver maintainer now. I'd prefer it to be the old way, but its not the end of the world. > > > > > > ndev->msix_entries = kmalloc(sizeof(struct msix_entry) * msix_entries, > > > @@ -1112,26 +1106,21 @@ static int ntb_setup_msix(struct ntb_device *ndev) > > > for (i = 0; i < msix_entries; i++) > > > ndev->msix_entries[i].entry = i; > > > > > > - rc = pci_enable_msix(pdev, ndev->msix_entries, msix_entries); > > > - if (rc < 0) > > > - goto err1; > > > - if (rc > 0) { > > > + if (ndev->hw_type != BWD_HW) > > > /* On SNB, the link interrupt is always tied to 4th vector. If > > > * we can't get all 4, then we can't use MSI-X. > > > */ > > > - if (ndev->hw_type != BWD_HW) { > > > - rc = -EIO; > > > - goto err1; > > > - } > > > - > > > - dev_warn(&pdev->dev, > > > - "Only %d MSI-X vectors. Limiting the number of queues to that number.\n", > > > - rc); > > > + rc = pci_enable_msix_range(pdev, ndev->msix_entries, > > > + msix_entries, msix_entries); > > > + else > > > + rc = pci_enable_msix_range(pdev, ndev->msix_entries, > > > + 1, msix_entries); > > > > Actually, this must be 2 for the min. One for the Data and one for > > the Link. If you look a few lines after this in the original code, > > there is a grabbing of the last vector for the link. I realize there > > is currently no check for this in the driver and a potential error > > case occurs. I can make a separate patch to correct this issue if > > this patch is not going through my tree. > > Well, I can re-post with a separate fix with this condition updated... > > } else if (rc < 2 || rc > ndev->limits.msix_cnt) { Won't this break for rc == 0 (the most probable case)? I'm fine with the fix being included in this patch, with a comment to note the fix. I don't believe it could ever be hit in real life, so I'm not worried about it. > rc = -EINVAL; > goto err; > } > > ...and the original patch with call to pci_enable_msix_range( ..., > 2, ndev->limits.msix_cnt) > > Will it work with both BWD_HW and non-BWD_HW hardware? non-BWD design has a min/max of 4...so it'll always have > 2 or be an error (hence the comment above). Thanks, Jon > > Thanks, > > Jon > > > > > > > + if (rc < 0) > > > + goto err1; > > > + else if (rc < msix_entries) { > > > + dev_warn(&pdev->dev, "Only %d MSI-X vectors. " > > > + "Limiting the number of queues to that number.\n", rc); > > > msix_entries = rc; > > > - > > > - rc = pci_enable_msix(pdev, ndev->msix_entries, msix_entries); > > > - if (rc) > > > - goto err1; > > > } > > > > > > for (i = 0; i < msix_entries; i++) { > > > diff --git a/drivers/ntb/ntb_hw.h b/drivers/ntb/ntb_hw.h > > > index bbdb7ed..d307107 100644 > > > --- a/drivers/ntb/ntb_hw.h > > > +++ b/drivers/ntb/ntb_hw.h > > > @@ -60,8 +60,6 @@ > > > #define PCI_DEVICE_ID_INTEL_NTB_SS_HSX 0x2F0F > > > #define PCI_DEVICE_ID_INTEL_NTB_B2B_BWD 0x0C4E > > > > > > -#define msix_table_size(control) ((control & PCI_MSIX_FLAGS_QSIZE)+1) > > > - > > > #ifndef readq > > > static inline u64 readq(void __iomem *addr) > > > { > > > -- > > > 1.7.7.6 > > > > > -- > Regards, > Alexander Gordeev > agordeev@redhat.com ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] ntb: Use pci_enable_msix_range() instead of pci_enable_msix() 2014-02-19 18:09 ` Jon Mason 2014-02-19 19:22 ` Alexander Gordeev @ 2014-02-20 13:22 ` Alexander Gordeev 1 sibling, 0 replies; 5+ messages in thread From: Alexander Gordeev @ 2014-02-20 13:22 UTC (permalink / raw) To: Jon Mason; +Cc: linux-kernel, linux-pci On Wed, Feb 19, 2014 at 11:09:27AM -0700, Jon Mason wrote: > On Wed, Feb 19, 2014 at 11:15:21AM +0100, Alexander Gordeev wrote: > > @@ -1112,26 +1106,21 @@ static int ntb_setup_msix(struct ntb_device *ndev) > > for (i = 0; i < msix_entries; i++) > > ndev->msix_entries[i].entry = i; > > > > - rc = pci_enable_msix(pdev, ndev->msix_entries, msix_entries); > > - if (rc < 0) > > - goto err1; > > - if (rc > 0) { > > + if (ndev->hw_type != BWD_HW) > > /* On SNB, the link interrupt is always tied to 4th vector. If > > * we can't get all 4, then we can't use MSI-X. > > */ > > - if (ndev->hw_type != BWD_HW) { > > - rc = -EIO; > > - goto err1; > > - } > > - > > - dev_warn(&pdev->dev, > > - "Only %d MSI-X vectors. Limiting the number of queues to that number.\n", > > - rc); > > + rc = pci_enable_msix_range(pdev, ndev->msix_entries, > > + msix_entries, msix_entries); > > + else > > + rc = pci_enable_msix_range(pdev, ndev->msix_entries, > > + 1, msix_entries); > > Actually, this must be 2 for the min. One for the Data and one for > the Link. If you look a few lines after this in the original code, > there is a grabbing of the last vector for the link. I realize there > is currently no check for this in the driver and a potential error > case occurs. I can make a separate patch to correct this issue if > this patch is not going through my tree. I am confused here again. In the original code I see "ntb-callback-msix" is requested for all MSI-Xs on BWD_HW and "ntb-event-msix" requested for the last vector on SNB_HW. Therefore, the range [1..msix_entries] should be fine for BWD_HW, isn't it? > Thanks, > Jon -- Regards, Alexander Gordeev agordeev@redhat.com ^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2014-02-20 13:20 UTC | newest] Thread overview: 5+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2014-02-19 10:15 [PATCH] ntb: Use pci_enable_msix_range() instead of pci_enable_msix() Alexander Gordeev 2014-02-19 18:09 ` Jon Mason 2014-02-19 19:22 ` Alexander Gordeev 2014-02-19 23:21 ` Jon Mason 2014-02-20 13:22 ` Alexander Gordeev
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).