QEMU-Devel Archive on lore.kernel.org
 help / color / Atom feed
From: Peter Xu <peterx@redhat.com>
To: Jason Wang <jasowang@redhat.com>
Cc: "Peter Maydell" <peter.maydell@linaro.org>,
	"Yan Zhao" <yan.y.zhao@intel.com>,
	"Juan Quintela" <quintela@redhat.com>,
	"Michael S. Tsirkin" <mst@redhat.com>,
	qemu-devel@nongnu.org, "Eugenio Pérez" <eperezma@redhat.com>,
	"Eric Auger" <eric.auger@redhat.com>,
	"Paolo Bonzini" <pbonzini@redhat.com>
Subject: Re: [RFC v2 1/1] memory: Delete assertion in memory_region_unregister_iommu_notifier
Date: Mon, 29 Jun 2020 09:34:03 -0400
Message-ID: <20200629133403.GA266532@xz-x1> (raw)
In-Reply-To: <54d2cdfd-97b8-9e1d-a607-d7a5e96be3a1@redhat.com>

On Mon, Jun 29, 2020 at 01:51:47PM +0800, Jason Wang wrote:
> 
> On 2020/6/28 下午10:47, Peter Xu wrote:
> > On Sun, Jun 28, 2020 at 03:03:41PM +0800, Jason Wang wrote:
> > > On 2020/6/27 上午5:29, Peter Xu wrote:
> > > > Hi, Eugenio,
> > > > 
> > > > (CCing Eric, Yan and Michael too)
> > > > 
> > > > On Fri, Jun 26, 2020 at 08:41:22AM +0200, Eugenio Pérez wrote:
> > > > > diff --git a/memory.c b/memory.c
> > > > > index 2f15a4b250..7f789710d2 100644
> > > > > --- a/memory.c
> > > > > +++ b/memory.c
> > > > > @@ -1915,8 +1915,6 @@ void memory_region_notify_one(IOMMUNotifier *notifier,
> > > > >            return;
> > > > >        }
> > > > > -    assert(entry->iova >= notifier->start && entry_end <= notifier->end);
> > > > I can understand removing the assertion should solve the issue, however imho
> > > > the major issue is not about this single assertion but the whole addr_mask
> > > > issue behind with virtio...
> > > 
> > > I don't get here, it looks to the the range was from guest IOMMU drivers.
> > Yes.  Note that I didn't mean that it's a problem in virtio, it's just the fact
> > that virtio is the only one I know that would like to support arbitrary address
> > range for the translated region.  I don't know about tcg, but vfio should still
> > need some kind of page alignment in both the address and the addr_mask.  We
> > have that assumption too across the memory core when we do translations.
> 
> 
> Right but it looks to me the issue is not the alignment.
> 
> 
> > 
> > A further cause of the issue is the MSI region when vIOMMU enabled - currently
> > we implemented the interrupt region using another memory region so it split the
> > whole DMA region into two parts.  That's really a clean approach to IR
> > implementation, however that's also a burden to the invalidation part because
> > then we'll need to handle things like this when the listened range is not page
> > alighed at all (neither 0-0xfedffff, nor 0xfef0000-MAX).  If without the IR
> > region (so the whole iommu address range will be a single FlatRange),
> 
> 
> Is this a bug? I remember that at least for vtd, it won't do any DMAR on the
> intrrupt address range

I don't think it's a bug, at least it's working as how I understand...  that
interrupt range is using an IR region, that's why I said the IR region splits
the DMAR region into two pieces, so we have two FlatRange for the same
IOMMUMemoryRegion.

> 
> 
> >   I think
> > we probably don't need most of the logic in vtd_address_space_unmap() at all,
> > then we can directly deliver all the IOTLB invalidations without splitting into
> > small page aligned ranges to all the iommu notifiers.  Sadly, so far I still
> > don't have ideal solution for it, because we definitely need IR.
> 
> 
> Another possible (theoretical) issue (for vhost) is that it can't trigger
> interrupt through the interrupt range.

Hmm.. Could you explain?  When IR is enabled, all devices including virtio
who send interrupt to 0xfeeXXXXX should be trapped by IR.

> 
> 
> > 
> > > 
> > > > For normal IOTLB invalidations, we were trying our best to always make
> > > > IOMMUTLBEntry contain a valid addr_mask to be 2**N-1.  E.g., that's what we're
> > > > doing with the loop in vtd_address_space_unmap().
> > > 
> > > I'm sure such such assumption can work for any type of IOMMU.
> > > 
> > > 
> > > > But this is not the first time that we may want to break this assumption for
> > > > virtio so that we make the IOTLB a tuple of (start, len), then that len can be
> > > > not a address mask any more.  That seems to be more efficient for things like
> > > > vhost because iotlbs there are not page based, so it'll be inefficient if we
> > > > always guarantee the addr_mask because it'll be quite a lot more roundtrips of
> > > > the same range of invalidation.  Here we've encountered another issue of
> > > > triggering the assertion with virtio-net, but only with the old RHEL7 guest.
> > > > 
> > > > I'm thinking whether we can make the IOTLB invalidation configurable by
> > > > specifying whether the backend of the notifier can handle arbitary address
> > > > range in some way.  So we still have the guaranteed addr_masks by default
> > > > (since I still don't think totally break the addr_mask restriction is wise...),
> > > > however we can allow the special backends to take adavantage of using arbitary
> > > > (start, len) ranges for reasons like performance.
> > > > 
> > > > To do that, a quick idea is to introduce a flag IOMMU_NOTIFIER_ARBITRARY_MASK
> > > > to IOMMUNotifierFlag, to declare that the iommu notifier (and its backend) can
> > > > take arbitrary address mask, then it can be any value and finally becomes a
> > > > length rather than an addr_mask.  Then for every iommu notify() we can directly
> > > > deliver whatever we've got from the upper layer to this notifier.  With the new
> > > > flag, vhost can do iommu_notifier_init() with UNMAP|ARBITRARY_MASK so it
> > > > declares this capability.  Then no matter for device iotlb or normal iotlb, we
> > > > skip the complicated procedure to split a big range into small ranges that are
> > > > with strict addr_mask, but directly deliver the message to the iommu notifier.
> > > > E.g., we can skip the loop in vtd_address_space_unmap() if the notifier is with
> > > > ARBITRARY flag set.
> > > 
> > > I'm not sure coupling IOMMU capability to notifier is the best choice.
> > IMHO it's not an IOMMU capability.  The flag I wanted to introduce is a
> > capability of the one who listens to the IOMMU TLB updates.  For our case, it's
> > virtio/vhost's capability to allow arbitrary length. The IOMMU itself
> > definitely has some limitation on the address range to be bound to an IOTLB
> > invalidation, e.g., the device-iotlb we're talking here only accept both the
> > iova address and addr_mask to be aligned to 2**N-1.
> 
> 
> I think this go back to one of our previous discussion of whether to
> introduce a dedicated notifiers for device IOTLB.
> 
> For IOMMU, it might have limitation like GAW, but for device IOTLB it
> probably doesn't. That's the reason we hit the assert here.

I feel like even for hardware it shouldn't be arbitrary either, because the
device iotlb sent from at least vt-d driver is very restricted too (borrowing
the comment you wrote :):

    /* According to ATS spec table 2.4:
     * S = 0, bits 15:12 = xxxx     range size: 4K
     * S = 1, bits 15:12 = xxx0     range size: 8K
     * S = 1, bits 15:12 = xx01     range size: 16K
     * S = 1, bits 15:12 = x011     range size: 32K
     * S = 1, bits 15:12 = 0111     range size: 64K
     * ...
     */

> 
> 
> > 
> > > How about just convert to use a range [start, end] for any notifier and move
> > > the checks (e.g the assert) into the actual notifier implemented (vhost or
> > > vfio)?
> > IOMMUTLBEntry itself is the abstraction layer of TLB entry.  Hardware TLB entry
> > is definitely not arbitrary range either (because AFAICT the hardware should
> > only cache PFN rather than address, so at least PAGE_SIZE aligned).
> > Introducing this flag will already make this trickier just to avoid introducing
> > another similar struct to IOMMUTLBEntry, but I really don't want to make it a
> > default option...  Not to mention I probably have no reason to urge the rest
> > iommu notifier users (tcg, vfio) to change their existing good code to suite
> > any of the backend who can cooperate with arbitrary address ranges...
> 
> 
> Ok, so it looks like we need a dedicated notifiers to device IOTLB.

Or we can also make a new flag for device iotlb just like current UNMAP? Then
we replace the vhost type from UNMAP to DEVICE_IOTLB.  But IMHO using the
ARBITRARY_LENGTH flag would work in a similar way.  DEVICE_IOTLB flag could
also allow virtio/vhost to only receive one invalidation (now IIUC it'll
receive both iotlb and device-iotlb for unmapping a page when ats=on), but then
ats=on will be a must and it could break some old (misconfiged) qemu because
afaict previously virtio/vhost could even work with vIOMMU (accidentally) even
without ats=on.

Thanks,

-- 
Peter Xu



  reply index

Thread overview: 48+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-06-26  6:41 [RFC v2 0/1] " Eugenio Pérez
2020-06-26  6:41 ` [RFC v2 1/1] " Eugenio Pérez
2020-06-26 21:29   ` Peter Xu
2020-06-27  7:26     ` Yan Zhao
2020-06-27 12:57       ` Peter Xu
2020-06-28  1:36         ` Yan Zhao
2020-06-28  7:03     ` Jason Wang
2020-06-28 14:47       ` Peter Xu
2020-06-29  5:51         ` Jason Wang
2020-06-29 13:34           ` Peter Xu [this message]
2020-06-30  2:41             ` Jason Wang
2020-06-30  8:29               ` Jason Wang
2020-06-30  9:21                 ` Michael S. Tsirkin
2020-06-30  9:23                   ` Jason Wang
2020-06-30 15:20                     ` Peter Xu
2020-07-01  8:11                       ` Jason Wang
2020-07-01 12:16                         ` Peter Xu
2020-07-01 12:30                           ` Jason Wang
2020-07-01 12:41                             ` Peter Xu
2020-07-02  3:00                               ` Jason Wang
2020-06-30 15:39               ` Peter Xu
2020-07-01  8:09                 ` Jason Wang
2020-07-02  3:01                   ` Jason Wang
2020-07-02 15:45                     ` Peter Xu
2020-07-03  7:24                       ` Jason Wang
2020-07-03 13:03                         ` Peter Xu
2020-07-07  8:03                           ` Jason Wang
2020-07-07 19:54                             ` Peter Xu
2020-07-08  5:42                               ` Jason Wang
2020-07-08 14:16                                 ` Peter Xu
2020-07-09  5:58                                   ` Jason Wang
2020-07-09 14:10                                     ` Peter Xu
2020-07-10  6:34                                       ` Jason Wang
2020-07-10 13:30                                         ` Peter Xu
2020-07-13  4:04                                           ` Jason Wang
2020-07-16  1:00                                             ` Peter Xu
2020-07-16  2:54                                               ` Jason Wang
2020-07-17 14:18                                                 ` Peter Xu
2020-07-20  4:02                                                   ` Jason Wang
2020-07-20 13:03                                                     ` Peter Xu
2020-07-21  6:20                                                       ` Jason Wang
2020-07-21 15:10                                                         ` Peter Xu
2020-08-03 16:00                         ` Eugenio Pérez
2020-08-04 20:30                           ` Peter Xu
2020-08-05  5:45                             ` Jason Wang
2020-06-29 15:05 ` [RFC v2 0/1] " Paolo Bonzini
2020-07-03  7:39   ` Eugenio Perez Martin
2020-07-03 10:10     ` Paolo Bonzini

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=20200629133403.GA266532@xz-x1 \
    --to=peterx@redhat.com \
    --cc=eperezma@redhat.com \
    --cc=eric.auger@redhat.com \
    --cc=jasowang@redhat.com \
    --cc=mst@redhat.com \
    --cc=pbonzini@redhat.com \
    --cc=peter.maydell@linaro.org \
    --cc=qemu-devel@nongnu.org \
    --cc=quintela@redhat.com \
    --cc=yan.y.zhao@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: link

QEMU-Devel Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/qemu-devel/0 qemu-devel/git/0.git
	git clone --mirror https://lore.kernel.org/qemu-devel/1 qemu-devel/git/1.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 qemu-devel qemu-devel/ https://lore.kernel.org/qemu-devel \
		qemu-devel@nongnu.org
	public-inbox-index qemu-devel

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.nongnu.qemu-devel


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git