From: Alex Williamson <alex.williamson@redhat.com>
To: linux-pci@vger.kernel.org, iommu@lists.linux-foundation.org
Cc: Joerg Roedel <joro@8bytes.org>,
linux-kernel@vger.kernel.org, acooks@gmail.com,
linux@horizon.com, bhelgaas@google.com, eddy0596@gmail.com
Subject: [PATCH v4 09/16] iommu/amd: Update to use PCI DMA aliases
Date: Thu, 22 May 2014 17:08:20 -0600 [thread overview]
Message-ID: <20140522230820.2856.67604.stgit@bling.home> (raw)
In-Reply-To: <20140522230230.2856.40017.stgit@bling.home>
AMD-Vi already has a concept of an alias provided via the IVRS table.
Now that PCI-core also understands aliases, we need to incorporate
both aspects when programming the IOMMU. IVRS is generally quite
reliable, so we continue to prefer it when an alias is present. For
cases where we have an IVRS alias that does not match the PCI alias
or where PCI does not report an alias, report the mismatch to allow
us to collect more quirks and dynamically incorporate the alias into
the device alias quirks where possible.
This should allow AMD-Vi to work with devices like Marvell and Ricoh
with DMA function alias quirks unknown to the BIOS.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Cc: Joerg Roedel <joro@8bytes.org>
---
drivers/iommu/amd_iommu.c | 78 +++++++++++++++++++++++++++++++++++++++++++--
1 file changed, 74 insertions(+), 4 deletions(-)
diff --git a/drivers/iommu/amd_iommu.c b/drivers/iommu/amd_iommu.c
index 57068e8..a73a7ed 100644
--- a/drivers/iommu/amd_iommu.c
+++ b/drivers/iommu/amd_iommu.c
@@ -427,6 +427,68 @@ use_group:
return use_dev_data_iommu_group(dev_data->alias_data, dev);
}
+static int __last_alias(struct pci_dev *pdev, u16 alias, void *data)
+{
+ *(u16 *)data = alias;
+ return 0;
+}
+
+static u16 get_alias(struct device *dev)
+{
+ struct pci_dev *pdev = to_pci_dev(dev);
+ u16 devid, ivrs_alias, pci_alias;
+
+ devid = get_device_id(dev);
+ ivrs_alias = amd_iommu_alias_table[devid];
+ pci_for_each_dma_alias(pdev, __last_alias, &pci_alias);
+
+ if (ivrs_alias == pci_alias)
+ return ivrs_alias;
+
+ /*
+ * DMA alias showdown
+ *
+ * The IVRS is fairly reliable in telling us about aliases, but it
+ * can't know about every screwy device. If we don't have an IVRS
+ * reported alias, use the PCI reported alias. In that case we may
+ * still need to initialize the rlookup and dev_table entries if the
+ * alias is to a non-existent device.
+ */
+ if (ivrs_alias == devid) {
+ if (!amd_iommu_rlookup_table[pci_alias]) {
+ amd_iommu_rlookup_table[pci_alias] =
+ amd_iommu_rlookup_table[devid];
+ memcpy(amd_iommu_dev_table[pci_alias].data,
+ amd_iommu_dev_table[devid].data,
+ sizeof(amd_iommu_dev_table[pci_alias].data));
+ }
+
+ return pci_alias;
+ }
+
+ pr_info("AMD-Vi: Using IVRS reported alias %02x:%02x.%d "
+ "for device %s[%04x:%04x], kernel reported alias "
+ "%02x:%02x.%d\n", PCI_BUS_NUM(ivrs_alias), PCI_SLOT(ivrs_alias),
+ PCI_FUNC(ivrs_alias), dev_name(dev), pdev->vendor, pdev->device,
+ PCI_BUS_NUM(pci_alias), PCI_SLOT(pci_alias),
+ PCI_FUNC(pci_alias));
+
+ /*
+ * If we don't have a PCI DMA alias and the IVRS alias is on the same
+ * bus, then the IVRS table may know about a quirk that we don't.
+ */
+ if (pci_alias == devid &&
+ PCI_BUS_NUM(ivrs_alias) == pdev->bus->number) {
+ pdev->dev_flags |= PCI_DEV_FLAGS_DMA_ALIAS_DEVFN;
+ pdev->dma_alias_devfn = ivrs_alias & 0xff;
+ pr_info("AMD-Vi: Added PCI DMA alias %02x.%d for %s\n",
+ PCI_SLOT(ivrs_alias), PCI_FUNC(ivrs_alias),
+ dev_name(dev));
+ }
+
+ return ivrs_alias;
+}
+
static int iommu_init_device(struct device *dev)
{
struct pci_dev *pdev = to_pci_dev(dev);
@@ -441,7 +503,8 @@ static int iommu_init_device(struct device *dev)
if (!dev_data)
return -ENOMEM;
- alias = amd_iommu_alias_table[dev_data->devid];
+ alias = get_alias(dev);
+
if (alias != dev_data->devid) {
struct iommu_dev_data *alias_data;
@@ -489,12 +552,19 @@ static void iommu_ignore_device(struct device *dev)
static void iommu_uninit_device(struct device *dev)
{
+ struct iommu_dev_data *dev_data = search_dev_data(get_device_id(dev));
+
+ if (!dev_data)
+ return;
+
iommu_group_remove_device(dev);
+ /* Unlink from alias, it may change if another device is re-plugged */
+ dev_data->alias_data = NULL;
+
/*
- * Nothing to do here - we keep dev_data around for unplugged devices
- * and reuse it when the device is re-plugged - not doing so would
- * introduce a ton of races.
+ * We keep dev_data around for unplugged devices and reuse it when the
+ * device is re-plugged - not doing so would introduce a ton of races.
*/
}
next prev parent reply other threads:[~2014-05-22 23:08 UTC|newest]
Thread overview: 32+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-05-22 23:07 [PATCH v4 00/16] PCI/iommu: Fix DMA alias problems Alex Williamson
2014-05-22 23:07 ` [PATCH v4 01/16] PCI: Add DMA alias iterator Alex Williamson
2014-05-22 23:07 ` [PATCH v4 02/16] PCI: define pci_dev_flags as bit shifts Alex Williamson
2014-05-22 23:07 ` [PATCH v4 03/16] PCI: quirk pci_for_each_dma_alias() Alex Williamson
2014-05-22 23:07 ` [PATCH v4 04/16] PCI: quirk dma_alias_devfn for Ricoh devices Alex Williamson
2014-05-22 23:07 ` [PATCH v4 05/16] PCI: quirk dma_alias_devfn for Marvell devices Alex Williamson
2014-05-23 1:29 ` George Spelvin
2014-05-28 17:55 ` Bjorn Helgaas
2014-05-28 18:04 ` Alex Williamson
2014-05-28 20:54 ` [PATCH v4.1 " Alex Williamson
2014-05-22 23:08 ` [PATCH v4 06/16] PCI: Quirk pci_for_each_dma_alias() for bridges Alex Williamson
2014-05-28 18:00 ` Bjorn Helgaas
2014-05-28 19:09 ` Alex Williamson
2014-05-28 20:57 ` [PATCH v4.1 " Alex Williamson
2014-05-22 23:08 ` [PATCH v4 07/16] PCI: Add quirks for ASMedia and Tundra bridges Alex Williamson
2014-05-22 23:08 ` [PATCH v4 08/16] iommu: Create central IOMMU group lookup/creation interface Alex Williamson
2014-05-22 23:08 ` Alex Williamson [this message]
2014-05-22 23:08 ` [PATCH v4 10/16] iommu/amd: Use iommu_group_get_for_dev() Alex Williamson
2014-05-22 23:08 ` [PATCH v4 11/16] iommu/intel: " Alex Williamson
2014-05-22 23:08 ` [PATCH v4 12/16] iommu/intel: Update to use PCI DMA aliases Alex Williamson
2014-05-22 23:08 ` [PATCH v4 13/16] iommu/fsl: Use iommu_group_get_for_dev() for IOMMU groups Alex Williamson
2014-05-30 9:17 ` Varun Sethi
2014-05-22 23:08 ` [PATCH v4 14/16] iommu: Remove pci.h Alex Williamson
2014-05-22 23:08 ` [PATCH v4 15/16] PCI: Remove pci_find_upstream_pcie_bridge() Alex Williamson
2014-05-22 23:09 ` [PATCH v4 16/16] PCI: Remove pci_get_dma_source() Alex Williamson
2014-05-28 5:23 ` [PATCH v4 00/16] PCI/iommu: Fix DMA alias problems Pat Erley
2014-05-28 20:29 ` Bjorn Helgaas
2014-05-28 20:45 ` Alex Williamson
2014-05-30 5:30 ` Andrew Cooks
2014-06-09 18:01 ` Alex Williamson
2014-06-16 14:47 ` Joerg Roedel
2014-06-16 15:34 ` Alex Williamson
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=20140522230820.2856.67604.stgit@bling.home \
--to=alex.williamson@redhat.com \
--cc=acooks@gmail.com \
--cc=bhelgaas@google.com \
--cc=eddy0596@gmail.com \
--cc=iommu@lists.linux-foundation.org \
--cc=joro@8bytes.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-pci@vger.kernel.org \
--cc=linux@horizon.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).