From: Yan Zhao <yan.y.zhao@intel.com>
To: Peter Xu <peterx@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>,
"qemu-devel@nongnu.org" <qemu-devel@nongnu.org>,
Auger Eric <eric.auger@redhat.com>
Subject: Re: [Qemu-devel] [PATCH] memory: do not do out of bound notification
Date: Mon, 24 Jun 2019 01:22:55 -0400 [thread overview]
Message-ID: <20190624052255.GA27894@joy-OptiPlex-7040> (raw)
In-Reply-To: <20190620130443.GC9657@xz-x1>
On Thu, Jun 20, 2019 at 09:04:43PM +0800, Peter Xu wrote:
> On Thu, Jun 20, 2019 at 08:59:55PM +0800, Peter Xu wrote:
> > On Thu, Jun 20, 2019 at 10:35:29AM +0200, Paolo Bonzini wrote:
> > > On 20/06/19 06:02, Peter Xu wrote:
> > > > Seems workable, to be explicit - we can even cut it into chunks with
> > > > different size to be efficient.
> > >
> > > Yes, this is not hard (completely untested):
> > >
> > > diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> > > index 44b1231157..541538bc6c 100644
> > > --- a/hw/i386/intel_iommu.c
> > > +++ b/hw/i386/intel_iommu.c
> > > @@ -3388,39 +3388,34 @@ static void vtd_address_space_unmap(VTDAddressSpace *as, IOMMUNotifier *n)
> > > }
> > >
> > > assert(start <= end);
> > > - size = end - start;
> > > + while (end > start) {
> > > + size = end - start;
> > > + /* Only keep the lowest bit of either size or start. */
> > > + size = MIN(size & -size, start & -start);
> >
> > I feel like this can be problematic. I'm imaging:
> >
> > start=0x1000_0000, size=0x1000_1000
> >
> > This will get size=0x1000 but actually we can do size=0x1000_0000 as
> > the first.
> >
> > > + /* Should not happen, but limit to address width too just in case */
> > > + size = MIN(size, 1ULL << s->aw_bits);
> > >
> > > - if (ctpop64(size) != 1) {
> > > - /*
> > > - * This size cannot format a correct mask. Let's enlarge it to
> > > - * suite the minimum available mask.
> > > - */
> > > - int n = 64 - clz64(size);
> > > - if (n > s->aw_bits) {
> > > - /* should not happen, but in case it happens, limit it */
> > > - n = s->aw_bits;
> > > - }
> > > - size = 1ULL << n;
> > > - }
> > > + assert((start & (size - 1)) == 0);
> > >
> > > - entry.target_as = &address_space_memory;
> > > - /* Adjust iova for the size */
> > > - entry.iova = n->start & ~(size - 1);
> > > - /* This field is meaningless for unmap */
> > > - entry.translated_addr = 0;
> > > - entry.perm = IOMMU_NONE;
> > > - entry.addr_mask = size - 1;
> > > + entry.target_as = &address_space_memory;
> > > + entry.iova = start;
> > > + /* This field is meaningless for unmap */
> > > + entry.translated_addr = 0;
> > > + entry.perm = IOMMU_NONE;
> > > + entry.addr_mask = size - 1;
> >
> > (some of the fields can be moved out of loop because they are
> > constants)
> >
> > >
> > > - trace_vtd_as_unmap_whole(pci_bus_num(as->bus),
> > > - VTD_PCI_SLOT(as->devfn),
> > > - VTD_PCI_FUNC(as->devfn),
> > > - entry.iova, size);
> > > + trace_vtd_as_unmap_whole(pci_bus_num(as->bus),
> > > + VTD_PCI_SLOT(as->devfn),
> > > + VTD_PCI_FUNC(as->devfn),
> > > + entry.iova, size);
> >
> > Can move this out because this is a trace only so we don't have
> > restriction on mask?
> >
> > >
> > > - map.iova = entry.iova;
> > > - map.size = entry.addr_mask;
> > > - iova_tree_remove(as->iova_tree, &map);
> > > + map.iova = entry.iova;
> > > + map.size = entry.addr_mask;
> > > + iova_tree_remove(as->iova_tree, &map);
> >
> > Same here?
> >
> > >
> > > - memory_region_notify_one(n, &entry);
> > > + memory_region_notify_one(n, &entry);
> > > + start += size;
> > > + }
> > > }
> > >
> > > static void vtd_address_space_unmap_all(IntelIOMMUState *s)
> > >
> > >
> > > Yan,
> > >
> > > if something like this works for you, let me know and I will submit it
> > > as a proper patch.
> > >
> > > Paolo
> >
> > Since during review I'm thinking how to generate a correct sequence of
> > these masks... here's my try below with above issues fixed... :)
> >
> > I've tried compile but not tested. Yan can test it, or I can do it
> > too tomorrow after I find some machines.
> >
> > Thanks,
> >
> > ------------------------------------------------------------
> > diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> > index 44b1231157..cfbd225f0a 100644
> > --- a/hw/i386/intel_iommu.c
> > +++ b/hw/i386/intel_iommu.c
> > @@ -3363,11 +3363,32 @@ VTDAddressSpace *vtd_find_add_as(IntelIOMMUState *s, PCIBus *bus, int devfn)
> > return vtd_dev_as;
> > }
> >
> > +static uint64_t vtd_get_next_mask(uint64_t start, uint64_t size, int gaw)
> > +{
> > + /* Tries to find smallest mask from start first */
> > + uint64_t rmask = start & -start, max_mask = 1ULL << gaw;
> > +
> > + assert(size && gaw > 0 && gaw < 64);
> > +
> > + /* Zero start, or too big */
> > + if (!rmask || rmask > max_mask) {
> > + rmask = max_mask;
> > + }
> > +
> > + /* If the start mask worked, then use it */
> > + if (rmask <= size) {
> > + return rmask;
> > + }
> > +
> > + /* Find the largest page mask from size */
> > + return 1ULL << (63 - clz64(size));
> > +}
> > +
> > /* Unmap the whole range in the notifier's scope. */
> > static void vtd_address_space_unmap(VTDAddressSpace *as, IOMMUNotifier *n)
> > {
> > IOMMUTLBEntry entry;
> > - hwaddr size;
> > + hwaddr size, remain;
> > hwaddr start = n->start;
> > hwaddr end = n->end;
> > IntelIOMMUState *s = as->iommu_state;
> > @@ -3388,39 +3409,28 @@ static void vtd_address_space_unmap(VTDAddressSpace *as, IOMMUNotifier *n)
> > }
> >
> > assert(start <= end);
> > - size = end - start;
> > -
> > - if (ctpop64(size) != 1) {
> > - /*
> > - * This size cannot format a correct mask. Let's enlarge it to
> > - * suite the minimum available mask.
> > - */
> > - int n = 64 - clz64(size);
> > - if (n > s->aw_bits) {
> > - /* should not happen, but in case it happens, limit it */
> > - n = s->aw_bits;
> > - }
> > - size = 1ULL << n;
> > - }
> > -
> > + size = remain = end - start;
> > entry.target_as = &address_space_memory;
> > - /* Adjust iova for the size */
> > - entry.iova = n->start & ~(size - 1);
> > + entry.perm = IOMMU_NONE;
> > /* This field is meaningless for unmap */
> > entry.translated_addr = 0;
> > - entry.perm = IOMMU_NONE;
> > - entry.addr_mask = size - 1;
> > +
> > + while (remain) {
> > + uint64_t mask = vtd_get_next_mask(start, remain, s->aw_bits);
> > +
> > + entry.iova = start;
> > + entry.addr_mask = mask - 1;
> > + memory_region_notify_one(n, &entry);
>
> Sorry, I at least missed these lines:
>
> start += mask;
> remain -= mask;
>
> > + }
> >
> > trace_vtd_as_unmap_whole(pci_bus_num(as->bus),
> > VTD_PCI_SLOT(as->devfn),
> > VTD_PCI_FUNC(as->devfn),
> > - entry.iova, size);
> > + n->start, size);
> >
> > - map.iova = entry.iova;
> > - map.size = entry.addr_mask;
> > + map.iova = n->start;
> > + map.size = size;
> > iova_tree_remove(as->iova_tree, &map);
> > -
> > - memory_region_notify_one(n, &entry);
> > }
> >
> > static void vtd_address_space_unmap_all(IntelIOMMUState *s)
> > ------------------------------------------------------------
> >
> > Regards,
> >
> > --
> > Peter Xu
>
> Regards,
>
> --
> Peter Xu
hi Peter and Paolo,
I tested with code and it's fine in my side.
It's base on your version with some minor modifications, such as size is
now (end - start + 1) now.
Thanks
Yan
+static uint64_t vtd_get_next_mask(uint64_t start, uint64_t size, int gaw)
+{
+ /* Tries to find smallest mask from start first */
+ uint64_t rmask = start & -start, max_mask = 1ULL << gaw;
+ assert(size && gaw > 0 && gaw < 64);
+ /* Zero start, or too big */
+ if (!rmask || rmask > max_mask) {
+ rmask = max_mask;
+ }
+ /* If the start mask worked, then use it */
+ if (rmask <= size) {
+ return rmask;
+ }
+
+ /* Find the largest page mask from size */
+ return 1ULL << (63 - clz64(size));
+}
+
/* Unmap the whole range in the notifier's scope. */
static void vtd_address_space_unmap(VTDAddressSpace *as, IOMMUNotifier *n)
{
IOMMUTLBEntry entry;
- hwaddr size;
+ hwaddr size, remain;
hwaddr start = n->start;
hwaddr end = n->end;
IntelIOMMUState *s = as->iommu_state;
@@ -3380,48 +3398,46 @@ static void vtd_address_space_unmap(VTDAddressSpace *as, IOMMUNotifier *n)
* VT-d spec), otherwise we need to consider overflow of 64 bits.
*/
- if (end > VTD_ADDRESS_SIZE(s->aw_bits)) {
+ if (end > VTD_ADDRESS_SIZE(s->aw_bits) - 1) {
/*
* Don't need to unmap regions that is bigger than the whole
* VT-d supported address space size
*/
- end = VTD_ADDRESS_SIZE(s->aw_bits);
+ end = VTD_ADDRESS_SIZE(s->aw_bits) - 1;
}
assert(start <= end);
- size = end - start;
- if (ctpop64(size) != 1) {
- /*
- * This size cannot format a correct mask. Let's enlarge it to
- * suite the minimum available mask.
- */
- int n = 64 - clz64(size);
- if (n > s->aw_bits) {
- /* should not happen, but in case it happens, limit it */
- n = s->aw_bits;
- }
- size = 1ULL << n;
- }
+ size = remain = end - start + 1;
entry.target_as = &address_space_memory;
- /* Adjust iova for the size */
- entry.iova = n->start & ~(size - 1);
+
+ entry.perm = IOMMU_NONE;
/* This field is meaningless for unmap */
entry.translated_addr = 0;
- entry.perm = IOMMU_NONE;
- entry.addr_mask = size - 1;
+
+ while (remain >= VTD_PAGE_SIZE) {
+ uint64_t mask = vtd_get_next_mask(start, remain, s->aw_bits);
+
+ entry.iova = start;
+ entry.addr_mask = mask - 1;
+ memory_region_notify_one(n, &entry);
+ start += mask;
+ remain -= mask;
+ }
+
+ if (remain) {
+ warn_report("Unmapping unaligned range %lx-%lx", start, end);
+ }
trace_vtd_as_unmap_whole(pci_bus_num(as->bus),
VTD_PCI_SLOT(as->devfn),
VTD_PCI_FUNC(as->devfn),
- entry.iova, size);
-
- map.iova = entry.iova;
- map.size = entry.addr_mask;
+ n->start, size);
+ map.iova = n->start;
+ map.size = size;
iova_tree_remove(as->iova_tree, &map);
- memory_region_notify_one(n, &entry);
}
static void vtd_address_space_unmap_all(IntelIOMMUState *s)
--
2.7.4
next prev parent reply other threads:[~2019-06-24 5:29 UTC|newest]
Thread overview: 17+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-06-19 8:49 [Qemu-devel] [PATCH] memory: do not do out of bound notification Yan Zhao
2019-06-19 13:17 ` Auger Eric
2019-06-20 1:46 ` Yan Zhao
2019-06-20 4:02 ` Peter Xu
2019-06-20 4:14 ` Yan Zhao
2019-06-20 8:14 ` Peter Xu
2019-06-20 8:13 ` Yan Zhao
2019-06-20 8:35 ` Paolo Bonzini
2019-06-20 10:57 ` Yan Zhao
2019-06-20 12:04 ` Paolo Bonzini
2019-06-20 12:59 ` Peter Xu
2019-06-20 13:04 ` Peter Xu
2019-06-24 5:22 ` Yan Zhao [this message]
2019-06-24 6:14 ` Peter Xu
2019-06-20 13:14 ` Paolo Bonzini
2019-06-21 2:36 ` Peter Xu
2019-06-21 7:57 ` Yan Zhao
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=20190624052255.GA27894@joy-OptiPlex-7040 \
--to=yan.y.zhao@intel.com \
--cc=eric.auger@redhat.com \
--cc=pbonzini@redhat.com \
--cc=peterx@redhat.com \
--cc=qemu-devel@nongnu.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).