ib: Use pci_enable_msix_range() instead of pci_enable_msix()

PATCH Superseded Alexander Gordeev 2 participants 7 messages Feb 18, 2014 – Apr 9, 2014

Subsystems (2)

PCI InfiniBand

lore

Share & tools

lore

Alexander Gordeev

author

2014-02-18 15:07 UTC #0 lore raw

cover ib: Use pci_enable_msix_range() instead of pci_enable_msix()

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.

Cc Roland Dreier <[email protected]>

Cc Sean Hefty <[email protected]>

Cc Hal Rosenstock <[email protected]>

Cc [email protected]

Alexander Gordeev (2):
mthca: Use pci_enable_msix_range() instead of pci_enable_msix()
qib: Use pci_enable_msix_range() instead of pci_enable_msix()

 drivers/infiniband/hw/mthca/mthca_main.c |    9 ++---
 drivers/infiniband/hw/qib/qib_pcie.c     |   55 +++++++++++++++--------------
 2 files changed, 31 insertions(+), 33 deletions(-)

Signature · 1 line

1.7.7.6

Alexander Gordeev

author

2014-02-18 15:07 UTC #1 lore raw

PATCH 1/2 mthca: Use pci_enable_msix_range() instead of pci_enable_msix()

Signed-off-by Alexander Gordeev <[email protected]>

Cc Roland Dreier <[email protected]>

Cc Sean Hefty <[email protected]>

Cc Hal Rosenstock <[email protected]>

Cc [email protected]

---
 drivers/infiniband/hw/mthca/mthca_main.c |    9 +++------
 1 files changed, 3 insertions(+), 6 deletions(-)

drivers/infiniband/hw/mthca/mthca_main.c +3 -6

index 87897b9..959d600 100644
@@ -858,13 +858,10 @@ static int mthca_enable_msi_x(struct mthca_dev *mdev)
 	entries[1].entry = 1;
 	entries[2].entry = 2;
 
-	err = pci_enable_msix(mdev->pdev, entries, ARRAY_SIZE(entries));
-	if (err) {
-		if (err > 0)
-			mthca_info(mdev, "Only %d MSI-X vectors available, "
-				   "not using MSI-X\n", err);
+	err = pci_enable_msix_range(mdev->pdev, entries,
+				    ARRAY_SIZE(entries), ARRAY_SIZE(entries));
+	if (err < 0)
 		return err;
-	}
 
 	mdev->eq_table.eq[MTHCA_EQ_COMP ].msi_x_vector = entries[0].vector;
 	mdev->eq_table.eq[MTHCA_EQ_ASYNC].msi_x_vector = entries[1].vector;

Signature · 1 line

1.7.7.6

Alexander Gordeev

author

2014-02-23 06:55 UTC #2 lore raw

PATCH v2 1/2 IB/mthca: Use pci_enable_msix_exact() instead of pci_enable_msix()

Signed-off-by Alexander Gordeev <[email protected]>

Cc Roland Dreier <[email protected]>

Cc Sean Hefty <[email protected]>

Cc Hal Rosenstock <[email protected]>

Cc [email protected]

---
 drivers/infiniband/hw/mthca/mthca_main.c |    8 ++------
 1 files changed, 2 insertions(+), 6 deletions(-)

drivers/infiniband/hw/mthca/mthca_main.c +2 -6

index 87897b9..ded76c1 100644
@@ -858,13 +858,9 @@ static int mthca_enable_msi_x(struct mthca_dev *mdev)
 	entries[1].entry = 1;
 	entries[2].entry = 2;
 
-	err = pci_enable_msix(mdev->pdev, entries, ARRAY_SIZE(entries));
-	if (err) {
-		if (err > 0)
-			mthca_info(mdev, "Only %d MSI-X vectors available, "
-				   "not using MSI-X\n", err);
+	err = pci_enable_msix_exact(mdev->pdev, entries, ARRAY_SIZE(entries));
+	if (err)
 		return err;
-	}
 
 	mdev->eq_table.eq[MTHCA_EQ_COMP ].msi_x_vector = entries[0].vector;
 	mdev->eq_table.eq[MTHCA_EQ_ASYNC].msi_x_vector = entries[1].vector;

Signature · 6 lines

1.7.7.6

-- 
Regards,
Alexander Gordeev
[email protected]

Alexander Gordeev

author

2014-02-18 15:07 UTC #3 lore raw

PATCH 2/2 qib: Use pci_enable_msix_range() instead of pci_enable_msix()

In reply to Alexander Gordeev

Signed-off-by Alexander Gordeev <[email protected]>

Cc Mike Marciniszyn <[email protected]>

Cc Roland Dreier <[email protected]>

Cc Sean Hefty <[email protected]>

Cc Hal Rosenstock <[email protected]>

Cc [email protected]

---
 drivers/infiniband/hw/qib/qib_pcie.c |   55 +++++++++++++++++----------------
 1 files changed, 28 insertions(+), 27 deletions(-)

drivers/infiniband/hw/qib/qib_pcie.c +28 ~~-27~~

index c8d9c4a..61a0046 100644
@@ -197,46 +197,47 @@ static void qib_msix_setup(struct qib_devdata *dd, int pos, u32 *msixcnt,
 			   struct qib_msix_entry *qib_msix_entry)
 {
 	int ret;
-	u32 tabsize = 0;
-	u16 msix_flags;
+	int nvec = *msixcnt;
 	struct msix_entry *msix_entry;
 	int i;
 
+	ret = pci_msix_vec_count(dd->pcidev);
+	if (ret < 0)
+		goto do_intx;
+
+	nvec = min(nvec, ret);
+
 	/* We can't pass qib_msix_entry array to qib_msix_setup
 	 * so use a dummy msix_entry array and copy the allocated
 	 * irq back to the qib_msix_entry array. */
-	msix_entry = kmalloc(*msixcnt * sizeof(*msix_entry), GFP_KERNEL);
-	if (!msix_entry) {
-		ret = -ENOMEM;
+	msix_entry = kmalloc(nvec * sizeof(*msix_entry), GFP_KERNEL);
+	if (!msix_entry)
 		goto do_intx;
-	}
-	for (i = 0; i < *msixcnt; i++)
+
+	for (i = 0; i < nvec; i++)
 		msix_entry[i] = qib_msix_entry[i].msix;
 
-	pci_read_config_word(dd->pcidev, pos + PCI_MSIX_FLAGS, &msix_flags);
-	tabsize = 1 + (msix_flags & PCI_MSIX_FLAGS_QSIZE);
-	if (tabsize > *msixcnt)
-		tabsize = *msixcnt;
-	ret = pci_enable_msix(dd->pcidev, msix_entry, tabsize);
-	if (ret > 0) {
-		tabsize = ret;
-		ret = pci_enable_msix(dd->pcidev, msix_entry, tabsize);
-	}
-do_intx:
-	if (ret) {
-		qib_dev_err(dd,
-			"pci_enable_msix %d vectors failed: %d, falling back to INTx\n",
-			tabsize, ret);
-		tabsize = 0;
-	}
-	for (i = 0; i < tabsize; i++)
+	ret = pci_enable_msix_range(dd->pcidev, msix_entry, 1, nvec);
+	if (ret < 0)
+		goto free_msix_entry;
+	else
+		nvec = ret;
+
+	for (i = 0; i < nvec; i++)
 		qib_msix_entry[i].msix = msix_entry[i];
+
 	kfree(msix_entry);
-	*msixcnt = tabsize;
+	*msixcnt = nvec;
+	return;
 
-	if (ret)
-		qib_enable_intx(dd->pcidev);
+free_msix_entry:
+	kfree(msix_entry);
 
+do_intx:
+	qib_dev_err(dd, "pci_enable_msix_range %d vectors failed: %d, "
+			"falling back to INTx\n", nvec, ret);
+	*msixcnt = 0;
+	qib_enable_intx(dd->pcidev);
 }
 
 /**

Signature · 1 line

1.7.7.6

Marciniszyn, Mike

2014-02-18 16:52 UTC #4 lore raw

RE: [PATCH 2/2] qib: Use pci_enable_msix_range() instead of pci_enable_msix()

Quoted 3 lines

Subject: [PATCH 2/2] qib: Use pci_enable_msix_range() instead of pci_enable_msix()

We are testing this now.

I suggest a subject/summary adding the IB/:
IB/qib: Use pci_enable_msix_range() instead of pci_enable_msix()

Marciniszyn, Mike

Acked-by on v1 2/2

2014-02-19 23:11 UTC #5 lore raw

RE: [PATCH 2/2] qib: Use pci_enable_msix_range() instead of pci_enable_msix()

In reply to Alexander Gordeev

Quoted 9 lines

Subject: [PATCH 2/2] qib: Use pci_enable_msix_range() instead of pci_enable_msix()
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 <[email protected]>

This worked fine!

Roland, can you fix up the subject to change qib to IB/qib when you merge it?

Acked-by Mike Marciniszyn <[email protected]>

Alexander Gordeev

author

2014-04-09 13:20 UTC #6 lore raw

Re: [PATCH 2/2] qib: Use pci_enable_msix_range() instead of pci_enable_msix()

Quoted 8 lines · Mike

Quoted 8 lines
Subject: [PATCH 2/2] qib: Use pci_enable_msix_range() instead of pci_enable_msix()
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.

Quoted 7 lines

Quoted 1 line
Signed-off-by Alexander Gordeev <[email protected]>
This worked fine!
Roland, can you fix up the subject to change qib to IB/qib when you merge it?
Acked-by Mike Marciniszyn <[email protected]>

Hi Roland,

Do you want me to repost these patches?

Thanks!

Signature · 3 lines

Regards,
Alexander Gordeev
[email protected]