From: Lu Baolu <baolu.lu@linux.intel.com> To: Joerg Roedel <joro@8bytes.org> Cc: ashok.raj@intel.com, jacob.jun.pan@linux.intel.com, kevin.tian@intel.com, Sai Praneeth Prakhya <sai.praneeth.prakhya@intel.com>, iommu@lists.linux-foundation.org, linux-kernel@vger.kernel.org, Daniel Drake <drake@endlessm.com>, Derrick Jonathan <jonathan.derrick@intel.com>, Jerry Snitselaar <jsnitsel@redhat.com>, Robin Murphy <robin.murphy@arm.com>, Christoph Hellwig <hch@lst.de>, Lu Baolu <baolu.lu@linux.intel.com> Subject: [PATCH v2 2/6] iommu: Configure default domain with def_domain_type Date: Sat, 14 Mar 2020 09:07:01 +0800 [thread overview] Message-ID: <20200314010705.30711-3-baolu.lu@linux.intel.com> (raw) In-Reply-To: <20200314010705.30711-1-baolu.lu@linux.intel.com> With the def_domain_type introduced, iommu default domain framework is now able to determine a proper default domain for a group. Let's use this to configure a group's default domain. If unlikely a device requires a special default domain type other than that in use, iommu probe procedure will either allocate a new domain according to the specified domain type, or (if the group has other devices sitting in it) change the default domain. The vendor iommu driver which exposes the def_domain_type callback should guarantee that there're no multiple devices in a same group requires differnt types of default domain. Signed-off-by: Sai Praneeth Prakhya <sai.praneeth.prakhya@intel.com> Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com> --- drivers/iommu/iommu.c | 93 +++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 90 insertions(+), 3 deletions(-) diff --git a/drivers/iommu/iommu.c b/drivers/iommu/iommu.c index 3e3528436e0b..b6a7429d0b45 100644 --- a/drivers/iommu/iommu.c +++ b/drivers/iommu/iommu.c @@ -1361,6 +1361,79 @@ struct iommu_group *fsl_mc_device_group(struct device *dev) } EXPORT_SYMBOL_GPL(fsl_mc_device_group); +/** + * Changes the default domain of a group + * + * @group: The group for which the default domain should be changed + * @type: The new default domain type + * + * The caller should hold the group->mutex before call into this function. + * Returns 0 on success and error code on failure. + */ +static int iommu_group_change_def_domain(struct iommu_group *group, int type) +{ + struct group_device *grp_dev, *temp; + struct iommu_domain *new, *old; + const struct iommu_ops *ops; + int ret = 0; + + if ((type != IOMMU_DOMAIN_IDENTITY && type != IOMMU_DOMAIN_DMA) || + !group->default_domain || type == group->default_domain->type || + !group->default_domain->ops) + return -EINVAL; + + if (group->domain != group->default_domain) + return -EBUSY; + + iommu_group_ref_get(group); + old = group->default_domain; + ops = group->default_domain->ops; + + /* Allocate a new domain of requested type. */ + new = ops->domain_alloc(type); + if (!new) { + ret = -ENOMEM; + goto domain_out; + } + new->type = type; + new->ops = ops; + new->pgsize_bitmap = group->default_domain->pgsize_bitmap; + + group->default_domain = new; + group->domain = new; + list_for_each_entry_safe(grp_dev, temp, &group->devices, list) { + struct device *dev; + + dev = grp_dev->dev; + if (device_is_bound(dev)) { + ret = -EINVAL; + goto device_out; + } + + iommu_group_create_direct_mappings(group, dev); + ret = __iommu_attach_device(group->domain, dev); + if (ret) + goto device_out; + + ret = ops->add_device(dev); + if (ret) + goto device_out; + } + + iommu_group_put(group); + iommu_domain_free(old); + + return 0; + +device_out: + group->default_domain = old; + __iommu_attach_group(old, group); + iommu_domain_free(new); +domain_out: + iommu_group_put(group); + return ret; +} + /** * iommu_group_get_for_dev - Find or create the IOMMU group for a device * @dev: target device @@ -1375,6 +1448,7 @@ struct iommu_group *iommu_group_get_for_dev(struct device *dev) { const struct iommu_ops *ops = dev->bus->iommu_ops; struct iommu_group *group; + int dev_def_type = 0; int ret; group = iommu_group_get(dev); @@ -1391,20 +1465,24 @@ struct iommu_group *iommu_group_get_for_dev(struct device *dev) if (IS_ERR(group)) return group; + if (ops->def_domain_type) + dev_def_type = ops->def_domain_type(dev); + /* * Try to allocate a default domain - needs support from the * IOMMU driver. */ if (!group->default_domain) { struct iommu_domain *dom; + int type = dev_def_type ? : iommu_def_domain_type; - dom = __iommu_domain_alloc(dev->bus, iommu_def_domain_type); - if (!dom && iommu_def_domain_type != IOMMU_DOMAIN_DMA) { + dom = __iommu_domain_alloc(dev->bus, type); + if (!dom && type != IOMMU_DOMAIN_DMA) { dom = __iommu_domain_alloc(dev->bus, IOMMU_DOMAIN_DMA); if (dom) { dev_warn(dev, "failed to allocate default IOMMU domain of type %u; falling back to IOMMU_DOMAIN_DMA", - iommu_def_domain_type); + type); } } @@ -1418,6 +1496,15 @@ struct iommu_group *iommu_group_get_for_dev(struct device *dev) DOMAIN_ATTR_DMA_USE_FLUSH_QUEUE, &attr); } + } else if (dev_def_type && + dev_def_type != group->default_domain->type) { + mutex_lock(&group->mutex); + ret = iommu_group_change_def_domain(group, dev_def_type); + mutex_unlock(&group->mutex); + if (ret) { + iommu_group_put(group); + return ERR_PTR(ret); + } } ret = iommu_group_add_device(group, dev); -- 2.17.1
WARNING: multiple messages have this Message-ID (diff)
From: Lu Baolu <baolu.lu@linux.intel.com> To: Joerg Roedel <joro@8bytes.org> Cc: kevin.tian@intel.com, ashok.raj@intel.com, linux-kernel@vger.kernel.org, Daniel Drake <drake@endlessm.com>, iommu@lists.linux-foundation.org, Robin Murphy <robin.murphy@arm.com>, Christoph Hellwig <hch@lst.de>, Derrick Jonathan <jonathan.derrick@intel.com> Subject: [PATCH v2 2/6] iommu: Configure default domain with def_domain_type Date: Sat, 14 Mar 2020 09:07:01 +0800 [thread overview] Message-ID: <20200314010705.30711-3-baolu.lu@linux.intel.com> (raw) In-Reply-To: <20200314010705.30711-1-baolu.lu@linux.intel.com> With the def_domain_type introduced, iommu default domain framework is now able to determine a proper default domain for a group. Let's use this to configure a group's default domain. If unlikely a device requires a special default domain type other than that in use, iommu probe procedure will either allocate a new domain according to the specified domain type, or (if the group has other devices sitting in it) change the default domain. The vendor iommu driver which exposes the def_domain_type callback should guarantee that there're no multiple devices in a same group requires differnt types of default domain. Signed-off-by: Sai Praneeth Prakhya <sai.praneeth.prakhya@intel.com> Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com> --- drivers/iommu/iommu.c | 93 +++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 90 insertions(+), 3 deletions(-) diff --git a/drivers/iommu/iommu.c b/drivers/iommu/iommu.c index 3e3528436e0b..b6a7429d0b45 100644 --- a/drivers/iommu/iommu.c +++ b/drivers/iommu/iommu.c @@ -1361,6 +1361,79 @@ struct iommu_group *fsl_mc_device_group(struct device *dev) } EXPORT_SYMBOL_GPL(fsl_mc_device_group); +/** + * Changes the default domain of a group + * + * @group: The group for which the default domain should be changed + * @type: The new default domain type + * + * The caller should hold the group->mutex before call into this function. + * Returns 0 on success and error code on failure. + */ +static int iommu_group_change_def_domain(struct iommu_group *group, int type) +{ + struct group_device *grp_dev, *temp; + struct iommu_domain *new, *old; + const struct iommu_ops *ops; + int ret = 0; + + if ((type != IOMMU_DOMAIN_IDENTITY && type != IOMMU_DOMAIN_DMA) || + !group->default_domain || type == group->default_domain->type || + !group->default_domain->ops) + return -EINVAL; + + if (group->domain != group->default_domain) + return -EBUSY; + + iommu_group_ref_get(group); + old = group->default_domain; + ops = group->default_domain->ops; + + /* Allocate a new domain of requested type. */ + new = ops->domain_alloc(type); + if (!new) { + ret = -ENOMEM; + goto domain_out; + } + new->type = type; + new->ops = ops; + new->pgsize_bitmap = group->default_domain->pgsize_bitmap; + + group->default_domain = new; + group->domain = new; + list_for_each_entry_safe(grp_dev, temp, &group->devices, list) { + struct device *dev; + + dev = grp_dev->dev; + if (device_is_bound(dev)) { + ret = -EINVAL; + goto device_out; + } + + iommu_group_create_direct_mappings(group, dev); + ret = __iommu_attach_device(group->domain, dev); + if (ret) + goto device_out; + + ret = ops->add_device(dev); + if (ret) + goto device_out; + } + + iommu_group_put(group); + iommu_domain_free(old); + + return 0; + +device_out: + group->default_domain = old; + __iommu_attach_group(old, group); + iommu_domain_free(new); +domain_out: + iommu_group_put(group); + return ret; +} + /** * iommu_group_get_for_dev - Find or create the IOMMU group for a device * @dev: target device @@ -1375,6 +1448,7 @@ struct iommu_group *iommu_group_get_for_dev(struct device *dev) { const struct iommu_ops *ops = dev->bus->iommu_ops; struct iommu_group *group; + int dev_def_type = 0; int ret; group = iommu_group_get(dev); @@ -1391,20 +1465,24 @@ struct iommu_group *iommu_group_get_for_dev(struct device *dev) if (IS_ERR(group)) return group; + if (ops->def_domain_type) + dev_def_type = ops->def_domain_type(dev); + /* * Try to allocate a default domain - needs support from the * IOMMU driver. */ if (!group->default_domain) { struct iommu_domain *dom; + int type = dev_def_type ? : iommu_def_domain_type; - dom = __iommu_domain_alloc(dev->bus, iommu_def_domain_type); - if (!dom && iommu_def_domain_type != IOMMU_DOMAIN_DMA) { + dom = __iommu_domain_alloc(dev->bus, type); + if (!dom && type != IOMMU_DOMAIN_DMA) { dom = __iommu_domain_alloc(dev->bus, IOMMU_DOMAIN_DMA); if (dom) { dev_warn(dev, "failed to allocate default IOMMU domain of type %u; falling back to IOMMU_DOMAIN_DMA", - iommu_def_domain_type); + type); } } @@ -1418,6 +1496,15 @@ struct iommu_group *iommu_group_get_for_dev(struct device *dev) DOMAIN_ATTR_DMA_USE_FLUSH_QUEUE, &attr); } + } else if (dev_def_type && + dev_def_type != group->default_domain->type) { + mutex_lock(&group->mutex); + ret = iommu_group_change_def_domain(group, dev_def_type); + mutex_unlock(&group->mutex); + if (ret) { + iommu_group_put(group); + return ERR_PTR(ret); + } } ret = iommu_group_add_device(group, dev); -- 2.17.1 _______________________________________________ iommu mailing list iommu@lists.linux-foundation.org https://lists.linuxfoundation.org/mailman/listinfo/iommu
next prev parent reply other threads:[~2020-03-14 1:09 UTC|newest] Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top 2020-03-14 1:06 [PATCH v2 0/6] Replace private domain with per-group default domain Lu Baolu 2020-03-14 1:06 ` Lu Baolu 2020-03-14 1:07 ` [PATCH v2 1/6] iommu: Add def_domain_type() callback in iommu_ops Lu Baolu 2020-03-14 1:07 ` Lu Baolu 2020-03-14 1:07 ` Lu Baolu [this message] 2020-03-14 1:07 ` [PATCH v2 2/6] iommu: Configure default domain with def_domain_type Lu Baolu 2020-03-19 14:03 ` Joerg Roedel 2020-03-19 14:03 ` Joerg Roedel 2020-03-20 9:06 ` Lu Baolu 2020-03-20 9:06 ` Lu Baolu 2020-03-14 1:07 ` [PATCH v2 3/6] iommu/vt-d: Don't force 32bit devices to uses DMA domain Lu Baolu 2020-03-14 1:07 ` Lu Baolu 2020-03-14 1:07 ` [PATCH v2 4/6] iommu/vt-d: Don't force PCI sub-hierarchy to use " Lu Baolu 2020-03-14 1:07 ` Lu Baolu 2020-03-14 1:07 ` [PATCH v2 5/6] iommu/vt-d: Add def_domain_type callback Lu Baolu 2020-03-14 1:07 ` Lu Baolu 2020-03-14 1:07 ` [PATCH v2 6/6] iommu/vt-d: Apply per-device dma_ops Lu Baolu 2020-03-14 1:07 ` Lu Baolu
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=20200314010705.30711-3-baolu.lu@linux.intel.com \ --to=baolu.lu@linux.intel.com \ --cc=ashok.raj@intel.com \ --cc=drake@endlessm.com \ --cc=hch@lst.de \ --cc=iommu@lists.linux-foundation.org \ --cc=jacob.jun.pan@linux.intel.com \ --cc=jonathan.derrick@intel.com \ --cc=joro@8bytes.org \ --cc=jsnitsel@redhat.com \ --cc=kevin.tian@intel.com \ --cc=linux-kernel@vger.kernel.org \ --cc=robin.murphy@arm.com \ --cc=sai.praneeth.prakhya@intel.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: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.