From: "Tian, Kevin" <kevin.tian-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org> To: "Liu, Yi L" <yi.l.liu-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>, Jean-Philippe Brucker <jean-philippe.brucker-5wv7dgnIgG8@public.gmane.org> Cc: "Lan, Tianyu" <tianyu.lan-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>, "kvm-u79uwXL29TY76Z2rM5mHXA@public.gmane.org" <kvm-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>, "jasowang-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org" <jasowang-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>, Will Deacon <Will.Deacon-5wv7dgnIgG8@public.gmane.org>, "qemu-devel-qX2TKyscuCcdnm+yROfE0A@public.gmane.org" <qemu-devel-qX2TKyscuCcdnm+yROfE0A@public.gmane.org>, "iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org" <iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org>, "Pan, Jacob jun" <jacob.jun.pan-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org> Subject: RE: [Qemu-devel] [RFC PATCH 7/8] VFIO: Add new IOCTL for IOMMU TLB invalidate propagation Date: Wed, 5 Jul 2017 06:45:38 +0000 [thread overview] Message-ID: <AADFC41AFE54684AB9EE6CBC0274A5D190D25919@SHSMSX101.ccr.corp.intel.com> (raw) In-Reply-To: <20170703103115.GB22053-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> > From: Liu, Yi L > Sent: Monday, July 3, 2017 6:31 PM > > Hi Jean, > > > > > > > 2. Define a structure in include/uapi/linux/iommu.h(newly added header > file) > > > > > > struct iommu_tlb_invalidate { > > > __u32 scope; > > > /* pasid-selective invalidation described by @pasid */ > > > #define IOMMU_INVALIDATE_PASID (1 << 0) > > > /* address-selevtive invalidation described by (@vaddr, @size) */ > > > #define IOMMU_INVALIDATE_VADDR (1 << 1) For VT-d above two flags are related. There is no method of flushing (@vaddr, @size) for all pasids, which doesn't make sense. address- selective invalidation is valid only for a given pasid. So it's not appropriate to put them in same level of scope definition at least for VT-d. > > > __u32 flags; > > > /* targets non-pasid mappings, @pasid is not valid */ > > > #define IOMMU_INVALIDATE_NO_PASID (1 << 0) > > > > Although it was my proposal, I don't like this flag. In ARM SMMU, we're > > using a special mode where PASID 0 is reserved and any traffic without > > PASID uses entry 0 of the PASID table. So I proposed the "NO_PASID" flag > > to invalidate that special context explicitly. But this means that > > invalidation packet targeted at that context will have "scope = PASID" and > > "flags = NO_PASID", which is utterly confusing. > > > > I now think that we should get rid of the IOMMU_INVALIDATE_NO_PASID > flag > > and just use PASID 0 to invalidate this context on ARM. I don't think > > other architectures would use the NO_PASID flag anyway, but might be > mistaken. > > I may suggest to keep it so far. On VT-d, we may pass some data in opaque, > so > we may work without it. But if other vendor want to issue non-PASID tagged > cache, then may encounter problem. I'm worried about what's the criteria which attribute should be abstracted in common structure and which can be left to opaque. It doesn't make much sense to do such abstraction purely because different vendor formats have some common fields. Usually we do such abstraction because vendor-agnostic code need to do some common handling before going to vendor specific code. However in this case VFIO is not expected to do anything with those IOMMU specific attributes. Then the structure is directly forwarded to IOMMU driver, which simply translates the structure into vendor specific opaque data again. Then why bothering to do double translations in Qemu and IOMMU driver side? Take VT-d for example. Below is a summary of all possible selections around invalidation of 1st level structure for svm: Scope: All PASIDs, single PASID for each PASID: all mappings, or page-selective mappings (addr, size) invalidation target: IOTLB entries (leaf) paging structure cache (non-leaf) PASID cache (pasid->cr3) invalidation hint: whether global pages are included drain reads/writes Above are pretty architectural attributes if just looking at functional purpose. Then if we really consider defining a common structure, it might be more natural to define a superset of all vendors' capabilities and remove the opaque field at all. But as said earlier the purpose of doing such abstraction is not clear if there is no vendor-agnostic user actually digesting those fields. Then should we reconsider the full opaque approach? Welcome comments since I may overlook something here. :-) Thanks Kevin
WARNING: multiple messages have this Message-ID (diff)
From: "Tian, Kevin" <kevin.tian@intel.com> To: "Liu, Yi L" <yi.l.liu@linux.intel.com>, Jean-Philippe Brucker <jean-philippe.brucker@arm.com> Cc: "Lan, Tianyu" <tianyu.lan@intel.com>, "Liu, Yi L" <yi.l.liu@intel.com>, "Raj, Ashok" <ashok.raj@intel.com>, "kvm@vger.kernel.org" <kvm@vger.kernel.org>, "jasowang@redhat.com" <jasowang@redhat.com>, Will Deacon <Will.Deacon@arm.com>, "alex.williamson@redhat.com" <alex.williamson@redhat.com>, "peterx@redhat.com" <peterx@redhat.com>, "qemu-devel@nongnu.org" <qemu-devel@nongnu.org>, "iommu@lists.linux-foundation.org" <iommu@lists.linux-foundation.org>, "Pan, Jacob jun" <jacob.jun.pan@intel.com> Subject: Re: [Qemu-devel] [RFC PATCH 7/8] VFIO: Add new IOCTL for IOMMU TLB invalidate propagation Date: Wed, 5 Jul 2017 06:45:38 +0000 [thread overview] Message-ID: <AADFC41AFE54684AB9EE6CBC0274A5D190D25919@SHSMSX101.ccr.corp.intel.com> (raw) In-Reply-To: <20170703103115.GB22053@gmail.com> > From: Liu, Yi L > Sent: Monday, July 3, 2017 6:31 PM > > Hi Jean, > > > > > > > 2. Define a structure in include/uapi/linux/iommu.h(newly added header > file) > > > > > > struct iommu_tlb_invalidate { > > > __u32 scope; > > > /* pasid-selective invalidation described by @pasid */ > > > #define IOMMU_INVALIDATE_PASID (1 << 0) > > > /* address-selevtive invalidation described by (@vaddr, @size) */ > > > #define IOMMU_INVALIDATE_VADDR (1 << 1) For VT-d above two flags are related. There is no method of flushing (@vaddr, @size) for all pasids, which doesn't make sense. address- selective invalidation is valid only for a given pasid. So it's not appropriate to put them in same level of scope definition at least for VT-d. > > > __u32 flags; > > > /* targets non-pasid mappings, @pasid is not valid */ > > > #define IOMMU_INVALIDATE_NO_PASID (1 << 0) > > > > Although it was my proposal, I don't like this flag. In ARM SMMU, we're > > using a special mode where PASID 0 is reserved and any traffic without > > PASID uses entry 0 of the PASID table. So I proposed the "NO_PASID" flag > > to invalidate that special context explicitly. But this means that > > invalidation packet targeted at that context will have "scope = PASID" and > > "flags = NO_PASID", which is utterly confusing. > > > > I now think that we should get rid of the IOMMU_INVALIDATE_NO_PASID > flag > > and just use PASID 0 to invalidate this context on ARM. I don't think > > other architectures would use the NO_PASID flag anyway, but might be > mistaken. > > I may suggest to keep it so far. On VT-d, we may pass some data in opaque, > so > we may work without it. But if other vendor want to issue non-PASID tagged > cache, then may encounter problem. I'm worried about what's the criteria which attribute should be abstracted in common structure and which can be left to opaque. It doesn't make much sense to do such abstraction purely because different vendor formats have some common fields. Usually we do such abstraction because vendor-agnostic code need to do some common handling before going to vendor specific code. However in this case VFIO is not expected to do anything with those IOMMU specific attributes. Then the structure is directly forwarded to IOMMU driver, which simply translates the structure into vendor specific opaque data again. Then why bothering to do double translations in Qemu and IOMMU driver side? Take VT-d for example. Below is a summary of all possible selections around invalidation of 1st level structure for svm: Scope: All PASIDs, single PASID for each PASID: all mappings, or page-selective mappings (addr, size) invalidation target: IOTLB entries (leaf) paging structure cache (non-leaf) PASID cache (pasid->cr3) invalidation hint: whether global pages are included drain reads/writes Above are pretty architectural attributes if just looking at functional purpose. Then if we really consider defining a common structure, it might be more natural to define a superset of all vendors' capabilities and remove the opaque field at all. But as said earlier the purpose of doing such abstraction is not clear if there is no vendor-agnostic user actually digesting those fields. Then should we reconsider the full opaque approach? Welcome comments since I may overlook something here. :-) Thanks Kevin
next prev parent reply other threads:[~2017-07-05 6:45 UTC|newest] Thread overview: 116+ messages / expand[flat|nested] mbox.gz Atom feed top 2017-04-26 10:11 [RFC PATCH 0/8] Shared Virtual Memory virtualization for VT-d Liu, Yi L 2017-04-26 10:11 ` [Qemu-devel] " Liu, Yi L 2017-04-26 10:11 ` [RFC PATCH 1/8] iommu: Introduce bind_pasid_table API function Liu, Yi L 2017-04-26 10:11 ` [Qemu-devel] " Liu, Yi L 2017-04-26 16:56 ` Jean-Philippe Brucker 2017-04-26 16:56 ` [Qemu-devel] " Jean-Philippe Brucker 2017-04-27 6:36 ` Liu, Yi L 2017-04-27 6:36 ` [Qemu-devel] " Liu, Yi L 2017-04-27 10:12 ` Jean-Philippe Brucker 2017-04-27 10:12 ` [Qemu-devel] " Jean-Philippe Brucker [not found] ` <772ca9de-50ba-a379-002d-5ff1f6a2e297-5wv7dgnIgG8@public.gmane.org> 2017-04-28 7:59 ` Liu, Yi L 2017-04-28 7:59 ` Liu, Yi L [not found] ` <c042bf90-d48b-4ebf-c01a-fca7c4875277-5wv7dgnIgG8@public.gmane.org> 2017-04-26 18:29 ` jacob pan 2017-04-26 18:29 ` [Qemu-devel] " jacob pan [not found] ` <20170426112948.00004520-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org> 2017-04-26 18:59 ` Jean-Philippe Brucker 2017-04-26 18:59 ` [Qemu-devel] " Jean-Philippe Brucker 2017-04-28 9:04 ` Liu, Yi L 2017-04-28 9:04 ` Liu, Yi L 2017-04-28 12:51 ` Jean-Philippe Brucker 2017-04-28 12:51 ` Jean-Philippe Brucker [not found] ` <3adb4e33-db96-4133-0510-412c3bfb24fe-5wv7dgnIgG8@public.gmane.org> 2017-05-23 7:50 ` Liu, Yi L 2017-05-23 7:50 ` Liu, Yi L 2017-05-25 12:33 ` Jean-Philippe Brucker 2017-05-25 12:33 ` Jean-Philippe Brucker [not found] ` <1493201525-14418-2-git-send-email-yi.l.liu-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org> 2017-05-12 21:59 ` Alex Williamson 2017-05-12 21:59 ` [Qemu-devel] " Alex Williamson [not found] ` <20170512155914.73bad777-1yVPhWWZRC1BDLzU/O5InQ@public.gmane.org> 2017-05-14 10:56 ` Liu, Yi L 2017-05-14 10:56 ` [Qemu-devel] " Liu, Yi L 2017-04-26 10:11 ` [RFC PATCH 2/8] iommu/vt-d: add bind_pasid_table function Liu, Yi L 2017-04-26 10:11 ` [Qemu-devel] " Liu, Yi L [not found] ` <1493201525-14418-3-git-send-email-yi.l.liu-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org> 2017-05-12 21:59 ` Alex Williamson 2017-05-12 21:59 ` [Qemu-devel] " Alex Williamson [not found] ` <20170512155929.66809113-1yVPhWWZRC1BDLzU/O5InQ@public.gmane.org> 2017-05-15 13:14 ` jacob pan 2017-05-15 13:14 ` [Qemu-devel] " jacob pan 2017-04-26 10:12 ` [RFC PATCH 3/8] iommu: Introduce iommu do invalidate API function Liu, Yi L 2017-04-26 10:12 ` [Qemu-devel] " Liu, Yi L [not found] ` <1493201525-14418-4-git-send-email-yi.l.liu-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org> 2017-05-12 21:59 ` Alex Williamson 2017-05-12 21:59 ` [Qemu-devel] " Alex Williamson [not found] ` <20170512155924.755ee17f-1yVPhWWZRC1BDLzU/O5InQ@public.gmane.org> 2017-05-17 10:23 ` Liu, Yi L 2017-05-17 10:23 ` [Qemu-devel] " Liu, Yi L 2017-04-26 10:12 ` [RFC PATCH 4/8] iommu/vt-d: Add iommu do invalidate function Liu, Yi L 2017-04-26 10:12 ` [Qemu-devel] " Liu, Yi L [not found] ` <1493201525-14418-5-git-send-email-yi.l.liu-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org> 2017-05-12 21:59 ` Alex Williamson 2017-05-12 21:59 ` [Qemu-devel] " Alex Williamson [not found] ` <20170512155918.5251fb94-1yVPhWWZRC1BDLzU/O5InQ@public.gmane.org> 2017-05-17 10:24 ` Liu, Yi L 2017-05-17 10:24 ` [Qemu-devel] " Liu, Yi L 2017-04-26 10:12 ` [RFC PATCH 5/8] VFIO: Add new IOTCL for PASID Table bind propagation Liu, Yi L 2017-04-26 10:12 ` [Qemu-devel] " Liu, Yi L 2017-04-26 16:56 ` Jean-Philippe Brucker 2017-04-26 16:56 ` [Qemu-devel] " Jean-Philippe Brucker 2017-04-27 5:43 ` Liu, Yi L [not found] ` <1493201525-14418-6-git-send-email-yi.l.liu-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org> 2017-05-11 10:29 ` Liu, Yi L 2017-05-11 10:29 ` Liu, Yi L 2017-05-12 21:58 ` Alex Williamson 2017-05-12 21:58 ` [Qemu-devel] " Alex Williamson [not found] ` <20170512155851.627409ed-1yVPhWWZRC1BDLzU/O5InQ@public.gmane.org> 2017-05-17 10:27 ` Liu, Yi L 2017-05-17 10:27 ` Liu, Yi L [not found] ` <20170517102759.GF22110-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> 2017-05-18 11:29 ` Jean-Philippe Brucker 2017-05-18 11:29 ` Jean-Philippe Brucker 2017-04-26 10:12 ` [RFC PATCH 6/8] VFIO: do pasid table binding Liu, Yi L 2017-04-26 10:12 ` [Qemu-devel] " Liu, Yi L 2017-05-09 7:55 ` Xiao Guangrong 2017-05-09 7:55 ` [Qemu-devel] " Xiao Guangrong 2017-05-11 10:29 ` Liu, Yi L 2017-05-12 21:59 ` Alex Williamson 2017-05-12 21:59 ` [Qemu-devel] " Alex Williamson 2017-04-26 10:12 ` [RFC PATCH 7/8] VFIO: Add new IOCTL for IOMMU TLB invalidate propagation Liu, Yi L 2017-04-26 10:12 ` [Qemu-devel] " Liu, Yi L 2017-05-12 12:11 ` Jean-Philippe Brucker 2017-05-12 12:11 ` [Qemu-devel] " Jean-Philippe Brucker [not found] ` <cc330a8f-e087-9b6f-2a40-38b58688d300-5wv7dgnIgG8@public.gmane.org> 2017-05-14 10:12 ` Liu, Yi L 2017-05-14 10:12 ` [Qemu-devel] " Liu, Yi L 2017-05-15 12:14 ` Jean-Philippe Brucker 2017-05-15 12:14 ` [Qemu-devel] " Jean-Philippe Brucker 2017-07-02 10:06 ` Liu, Yi L 2017-07-02 10:06 ` Liu, Yi L 2017-07-03 11:52 ` Jean-Philippe Brucker [not found] ` <0e4f2dd4-d553-b1b7-7bec-fe0ff5242c54-5wv7dgnIgG8@public.gmane.org> 2017-07-03 10:31 ` Liu, Yi L 2017-07-03 10:31 ` Liu, Yi L [not found] ` <20170703103115.GB22053-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> 2017-07-05 6:45 ` Tian, Kevin [this message] 2017-07-05 6:45 ` Tian, Kevin [not found] ` <AADFC41AFE54684AB9EE6CBC0274A5D190D25919-0J0gbvR4kThpB2pF5aRoyrfspsVTdybXVpNB7YpNyf8@public.gmane.org> 2017-07-05 12:42 ` Jean-Philippe Brucker 2017-07-05 12:42 ` Jean-Philippe Brucker [not found] ` <1d63c1ae-ca10-0f9d-91de-0d9c9823c104-5wv7dgnIgG8@public.gmane.org> 2017-07-05 17:28 ` Alex Williamson 2017-07-05 17:28 ` Alex Williamson [not found] ` <20170705112816.56554f65-DGNDKt5SQtizQB+pC5nmwQ@public.gmane.org> 2017-07-05 22:26 ` Tian, Kevin 2017-07-05 22:26 ` Tian, Kevin 2017-07-14 8:58 ` Liu, Yi L 2017-07-14 8:58 ` Liu, Yi L [not found] ` <A2975661238FB949B60364EF0F2C2574390A7C4F-E2R4CRU6q/6iAffOGbnezLfspsVTdybXVpNB7YpNyf8@public.gmane.org> 2017-07-14 18:15 ` Alex Williamson 2017-07-14 18:15 ` Alex Williamson [not found] ` <20170714121555.7e64d849-DGNDKt5SQtizQB+pC5nmwQ@public.gmane.org> 2017-07-17 10:58 ` Liu, Yi L 2017-07-17 10:58 ` Liu, Yi L 2017-07-17 22:45 ` Alex Williamson [not found] ` <20170717164515.2491b3bf-DGNDKt5SQtizQB+pC5nmwQ@public.gmane.org> 2017-07-18 9:38 ` Jean-Philippe Brucker 2017-07-18 9:38 ` Jean-Philippe Brucker [not found] ` <d0abeefc-adcf-85c3-f5d9-8c90a18f8011-5wv7dgnIgG8@public.gmane.org> 2017-07-18 14:29 ` Alex Williamson 2017-07-18 14:29 ` Alex Williamson 2017-07-18 15:03 ` Jean-Philippe Brucker 2017-07-19 10:45 ` Liu, Yi L 2017-07-19 10:45 ` Liu, Yi L 2017-07-19 21:50 ` Jacob Pan 2017-07-19 21:50 ` Jacob Pan 2017-07-05 22:31 ` Tian, Kevin 2017-07-05 22:31 ` Tian, Kevin 2017-05-12 21:58 ` Alex Williamson 2017-05-12 21:58 ` [Qemu-devel] " Alex Williamson 2017-05-14 10:55 ` Liu, Yi L 2017-05-14 10:55 ` [Qemu-devel] " Liu, Yi L [not found] ` <20170514105507.GB22110-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> 2017-07-05 5:32 ` Tian, Kevin 2017-07-05 5:32 ` [Qemu-devel] " Tian, Kevin 2017-04-26 10:12 ` [RFC PATCH 8/8] VFIO: do IOMMU TLB invalidation from guest Liu, Yi L 2017-04-26 10:12 ` [Qemu-devel] " Liu, Yi L 2017-05-08 4:09 ` [RFC PATCH 0/8] Shared Virtual Memory virtualization for VT-d Xiao Guangrong 2017-05-08 4:09 ` [Qemu-devel] " Xiao Guangrong 2017-05-07 7:33 ` Liu, Yi L
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=AADFC41AFE54684AB9EE6CBC0274A5D190D25919@SHSMSX101.ccr.corp.intel.com \ --to=kevin.tian-ral2jqcrhueavxtiumwx3w@public.gmane.org \ --cc=Will.Deacon-5wv7dgnIgG8@public.gmane.org \ --cc=iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org \ --cc=jacob.jun.pan-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org \ --cc=jasowang-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org \ --cc=jean-philippe.brucker-5wv7dgnIgG8@public.gmane.org \ --cc=kvm-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \ --cc=qemu-devel-qX2TKyscuCcdnm+yROfE0A@public.gmane.org \ --cc=tianyu.lan-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org \ --cc=yi.l.liu-VuQAYsv1563Yd54FQh9/CA@public.gmane.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: 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.