From: Dongdong Liu <liudongdong3@huawei.com>
To: <helgaas@kernel.org>, <hch@lst.de>
Cc: <linux-pci@vger.kernel.org>, <stable@vger.kernel.org>,
<gabriele.paoloni@huawei.com>, <charles.chenxin@huawei.com>,
<linuxarm@huawei.com>, Dongdong Liu <liudongdong3@huawei.com>
Subject: [PATCH V4 1/2] PCI/portdrv: Fix MSI/MSI-X bug for PCIe port service drivers
Date: Wed, 11 Oct 2017 18:52:57 +0800 [thread overview]
Message-ID: <1507719178-112556-2-git-send-email-liudongdong3@huawei.com> (raw)
In-Reply-To: <1507719178-112556-1-git-send-email-liudongdong3@huawei.com>
After removing and adding back the PCI root port device,
we see the PCIe port service drivers request irq failed.
pcie_pme: probe of 0000:00:00.0:pcie001 failed with error -22
aer: probe of 0000:00:00.0:pcie002 failed with error -22
pciehp 0000:00:00.0:pcie004: Slot #0 AttnBtn- PwrCtrl- MRL- AttnInd-
PwrInd- HotPlug+ Surprise+ Interlock- NoCompl- LLActRep+
pciehp 0000:00:00.0:pcie004: Cannot get irq 20 for the hotplug controller
pciehp 0000:00:00.0:pcie004: Notification initialization failed (-1)
dpc 0000:00:00.0:pcie010: request IRQ22 failed: -22
dpc: probe of 0000:00:00.0:pcie010 failed with error -22
The current code basically does this:
- allocate 32 vectors
- figure out vector used by PME and hotplug
- figure out vector used by AER
- figure out vector used by DPC
- free the 32 vectors we allocated
- allocate only as many vectors as we need
but it is broken as calling pci_free_irq_vectors()
invalidates the IRQ numbers returned before by pci_irq_vectors();
The hardware works:
- PME and hotplug use the Interrupt Message Number from the PCIe
Capability register.
- AER uses the AER Interrupt Message Number from the AER Root Error
Status register.
- DPC uses the DPC Interrupt Message Number from the DPC Capability
register.
- FRS (not supported by Linux yet) uses the FRS Interrupt Message
Number from the FRS Queuing Capability register.
- That's a total of 4 possible MSI/MSI-X vectors used for PME,
hotplug, AER, DPC, and FRS, so there's no point in trying to
allocate more than 4 vectors (we currently start with 32).
- All these Interrupt Message Numbers are read-only to software but
are updated by hardware when software writes the Multiple Message
Enable field of the MSI Message Control register. Thanks for
pointing this out; I didn't realize this before.
- Therefore, we must read the Interrupt Message Numbers *after* we
write Multiple Message Enable.
Since we can use at most 4 vectors, this patch is to optimize the
interrupt vector allocation strategy in pcie_port_enable_irq_vec().
First figure out how many vectors we need,allocate them, and then fill
in the slots in the irqs[] table.
Cc: <stable@vger.kernel.org>
Fixes: 3674cc49da9a ("PCI/portdrv: Use pci_irq_alloc_vectors()")
Suggested-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Dongdong Liu <liudongdong3@huawei.com>
Signed-off-by: Gabriele Paoloni <gabriele.paoloni@huawei.com>
---
drivers/pci/pcie/portdrv_core.c | 45 +++++++++++++----------------------------
1 file changed, 14 insertions(+), 31 deletions(-)
diff --git a/drivers/pci/pcie/portdrv_core.c b/drivers/pci/pcie/portdrv_core.c
index 313a21d..9692379 100644
--- a/drivers/pci/pcie/portdrv_core.c
+++ b/drivers/pci/pcie/portdrv_core.c
@@ -56,14 +56,16 @@ static int pcie_port_enable_irq_vec(struct pci_dev *dev, int *irqs, int mask)
{
int nr_entries, entry, nvec = 0;
- /*
- * Allocate as many entries as the port wants, so that we can check
- * which of them will be useful. Moreover, if nr_entries is correctly
- * equal to the number of entries this port actually uses, we'll happily
- * go through without any tricks.
- */
- nr_entries = pci_alloc_irq_vectors(dev, 1, PCIE_PORT_MAX_MSI_ENTRIES,
- PCI_IRQ_MSIX | PCI_IRQ_MSI);
+ if (mask & (PCIE_PORT_SERVICE_PME | PCIE_PORT_SERVICE_HP))
+ nvec++;
+ if (mask & PCIE_PORT_SERVICE_AER)
+ nvec++;
+ if (mask & PCIE_PORT_SERVICE_DPC)
+ nvec++;
+
+ /* Allocate as many vectors as the we need */
+ nr_entries = pci_alloc_irq_vectors(dev, 1, nvec,
+ PCI_IRQ_MSIX | PCI_IRQ_MSI);
if (nr_entries < 0)
return nr_entries;
@@ -90,10 +92,10 @@ static int pcie_port_enable_irq_vec(struct pci_dev *dev, int *irqs, int mask)
if (entry >= nr_entries)
goto out_free_irqs;
- irqs[PCIE_PORT_SERVICE_PME_SHIFT] = pci_irq_vector(dev, entry);
- irqs[PCIE_PORT_SERVICE_HP_SHIFT] = pci_irq_vector(dev, entry);
-
- nvec = max(nvec, entry + 1);
+ if (mask & PCIE_PORT_SERVICE_PME)
+ irqs[PCIE_PORT_SERVICE_PME_SHIFT] = pci_irq_vector(dev, entry);
+ if (mask & PCIE_PORT_SERVICE_HP)
+ irqs[PCIE_PORT_SERVICE_HP_SHIFT] = pci_irq_vector(dev, entry);
}
if (mask & PCIE_PORT_SERVICE_AER) {
@@ -120,7 +122,6 @@ static int pcie_port_enable_irq_vec(struct pci_dev *dev, int *irqs, int mask)
irqs[PCIE_PORT_SERVICE_AER_SHIFT] = pci_irq_vector(dev, entry);
- nvec = max(nvec, entry + 1);
}
if (mask & PCIE_PORT_SERVICE_DPC) {
@@ -146,24 +147,6 @@ static int pcie_port_enable_irq_vec(struct pci_dev *dev, int *irqs, int mask)
goto out_free_irqs;
irqs[PCIE_PORT_SERVICE_DPC_SHIFT] = pci_irq_vector(dev, entry);
-
- nvec = max(nvec, entry + 1);
- }
-
- /*
- * If nvec is equal to the allocated number of entries, we can just use
- * what we have. Otherwise, the port has some extra entries not for the
- * services we know and we need to work around that.
- */
- if (nvec != nr_entries) {
- /* Drop the temporary MSI-X setup */
- pci_free_irq_vectors(dev);
-
- /* Now allocate the MSI-X vectors for real */
- nr_entries = pci_alloc_irq_vectors(dev, nvec, nvec,
- PCI_IRQ_MSIX | PCI_IRQ_MSI);
- if (nr_entries < 0)
- return nr_entries;
}
return 0;
--
1.9.1
next prev parent reply other threads:[~2017-10-11 10:52 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-10-11 10:52 [PATCH V4 0/2] PCI/portdrv: Fix MSI/MSI-X bug for PCIe port service drivers Dongdong Liu
2017-10-11 10:52 ` Dongdong Liu [this message]
2017-10-11 11:30 ` [PATCH V4 1/2] " Bjorn Helgaas
2017-10-12 2:40 ` Dongdong Liu
2017-10-17 20:25 ` Bjorn Helgaas
2017-10-18 10:50 ` Dongdong Liu
2017-10-18 22:33 ` Bjorn Helgaas
2017-10-19 2:50 ` Dongdong Liu
2017-10-19 22:47 ` Bjorn Helgaas
2017-10-11 10:52 ` [PATCH V4 2/2] PCI/portdrv: Add #defines for the AER and DPC Interrupt Message Number masks Dongdong Liu
2017-10-11 13:01 ` David Laight
2017-10-11 13:34 ` Bjorn Helgaas
2017-10-16 12:09 ` [PATCH V4 0/2] PCI/portdrv: Fix MSI/MSI-X bug for PCIe port service drivers Dongdong Liu
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=1507719178-112556-2-git-send-email-liudongdong3@huawei.com \
--to=liudongdong3@huawei.com \
--cc=charles.chenxin@huawei.com \
--cc=gabriele.paoloni@huawei.com \
--cc=hch@lst.de \
--cc=helgaas@kernel.org \
--cc=linux-pci@vger.kernel.org \
--cc=linuxarm@huawei.com \
--cc=stable@vger.kernel.org \
/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).