Change for PCI core to use pci_is_pcie() instead of checking
pci_dev->is_pcie.
Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
To support PCIe hot plug in IOMMU, we register a notifier to respond to device
change action.
When the notifier gets BUS_NOTIFY_UNBOUND_DRIVER, it removes the device
from its DMAR domain.
A hot added device will be added into an IOMMU domain when it first does IOMMU
op. So there is no need to add more code for hot add.
Without the patch, after a hot-remove, a hot-added device on the same
slot will not work.
Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Tested-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
The model for IOMMU passthrough is that decent devices that can cope
with DMA to all of memory get passthrough; crappy devices with a limited
dma_mask don't -- they get to use the IOMMU anyway.
This is done on the basis that IOMMU passthrough is usually wanted for
performance reasons, and it's only the decent PCI devices that you
really care about performance for, while the crappy 32-bit ones like
your USB controller can just use the IOMMU and you won't really care.
Unfortunately, the check for this was only looking at dev->dma_mask, not
at dev->coherent_dma_mask. And some devices have a 32-bit
coherent_dma_mask even though they have a full 64-bit dma_mask.
Even more unfortunately, fixing that simple oversight would upset
certain broken HP devices. Not only do they have a 32-bit
coherent_dma_mask, but they also have a tendency to do stray DMA to
unmapped addresses. And then they die when they take the DMA fault they
so richly deserve.
So if we do the 'correct' fix, it'll mean that affected users have to
disable IOMMU support completely on "a large percentage of servers from
a major vendor."
Personally, I have little sympathy -- given that this is the _same_
'major vendor' who is shipping machines which claim to have IOMMU
support but have obviously never _once_ booted a VT-d capable OS to do
any form of QA. But strictly speaking, it _would_ be a regression even
though it only ever worked by fluke.
For 2.6.33, we'll come up with a quirk which gives swiotlb support
for this particular device, and other devices with an inadequate
coherent_dma_mask will just get normal IOMMU mapping.
The simplest fix for 2.6.32, though, is just to jump through some hoops
to try to allocate coherent DMA memory for such devices in a place that
they can reach. We'd use dma_generic_alloc_coherent() for this if it
existed on IA64.
Signed-off-by: Alex Williamson <alex.williamson@hp.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
If HW IOMMU initialization fails (Intel VT-d often does this,
typically due to BIOS bugs), we fall back to nommu. It doesn't
work for the majority since nowadays we have more than 4GB
memory so we must use swiotlb instead of nommu.
The problem is that it's too late to initialize swiotlb when HW
IOMMU initialization fails. We need to allocate swiotlb memory
earlier from bootmem allocator. Chris explained the issue in
detail:
http://marc.info/?l=linux-kernel&m=125657444317079&w=2
The current x86 IOMMU initialization sequence is too complicated
and handling the above issue makes it more hacky.
This patch changes x86 IOMMU initialization sequence to handle
the above issue cleanly.
The new x86 IOMMU initialization sequence are:
1. we initialize the swiotlb (and setting swiotlb to 1) in the case
of (max_pfn > MAX_DMA32_PFN && !no_iommu). dma_ops is set to
swiotlb_dma_ops or nommu_dma_ops. if swiotlb usage is forced by
the boot option, we finish here.
2. we call the detection functions of all the IOMMUs
3. the detection function sets x86_init.iommu.iommu_init to the
IOMMU initialization function (so we can avoid calling the
initialization functions of all the IOMMUs needlessly).
4. if the IOMMU initialization function doesn't need to swiotlb
then sets swiotlb to zero (e.g. the initialization is
sucessful).
5. if we find that swiotlb is set to zero, we free swiotlb
resource.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: chrisw@sous-sol.org
Cc: dwmw2@infradead.org
Cc: joerg.roedel@amd.com
Cc: muli@il.ibm.com
LKML-Reference: <1257849980-22640-10-git-send-email-fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Asus decided to ship a BIOS which configures sound DMA to go via the
dedicated IOMMU unit, but assigns precisely zero TLB entries to that
unit. Which causes the whole thing to deadlock, including the DMA
traffic on the _other_ IOMMU units. Nice one.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
The following 64 bit promotions are necessary to handle memory above the
4GiB boundary correctly.
[dwmw2: Fix the second part not to need 64-bit arithmetic at all]
Signed-off-by: Benjamin LaHaise <ben.lahaise@neterion.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
If end_pfn is equal to (unsigned long)-1, then the loop will never end.
Seen on 32-bit kernel, but could have happened on 64-bit too once we get
hardware that supports 64-bit guest addresses.
Change both functions to a 'do {} while' loop with the test at the end,
and check for the PFN having wrapper round to zero.
Reported-by: Benjamin LaHaise <ben.lahaise@neterion.com>
Tested-by: Benjamin LaHaise <ben.lahaise@neterion.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
This means we're limited to 44-bit addresses on 32-bit kernels, and
makes it sane for us to use 'unsigned long' for PFNs throughout.
Which is just as well, really, since we already do that.
Reported-by: Benjamin LaHaise <ben.lahaise@neterion.com>
Tested-by: Benjamin LaHaise <ben.lahaise@neterion.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Move tboot.h from asm to linux to fix the build errors of intel_txt
patch on non-X86 platforms. Remove the tboot code from generic code
init/main.c and kernel/cpu.c.
Signed-off-by: Shane Wang <shane.wang@intel.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
This file needs to include linux/dmi.h directly rather than relying on
it being pulled in from elsewhere.
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
The kcalloc() failure path in iommu_init_domains() calls
free_dmar_iommu(), which assumes that ->domains, ->domain_ids,
and ->lock have been properly initialized.
Add checks in free_[dmar]_iommu to not use ->domains,->domain_ids
if not alloced. Move the lock init to prior to the kcalloc()'s,
so it is valid in free_context_table() when free_dmar_iommu() invokes
it at the end.
Patch based on iommu-2.6,
commit 132032274a
Signed-off-by: Donald Dutile <ddutile@redhat.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Mark si_domain_init and iommu_prepare_static_identity_mapping with
__init, to eliminate the following warnings:
WARNING: drivers/pci/built-in.o(.text+0xf1f4): Section mismatch in reference from the function si_domain_init() to the function .init.text:si_domain_work_fn()
The function si_domain_init() references
the function __init si_domain_work_fn().
This is often because si_domain_init lacks a __init
annotation or the annotation of si_domain_work_fn is wrong.
WARNING: drivers/pci/built-in.o(.text+0xe340): Section mismatch in reference from the function iommu_prepare_static_identity_mapping() to the function .init.text:si_domain_init()
The function iommu_prepare_static_identity_mapping() references
the function __init si_domain_init().
This is often because iommu_prepare_static_identity_mapping lacks a __init
annotation or the annotation of si_domain_init is wrong.
Signed-off-by: Matt Kraai <kraai@ftbfs.org>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
All callers of the former were also calling the latter, in one order or
the other, and failing to correctly clean up if the second returned
failure.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Two defects work together result in KVM device passthrough randomly can't
work:
1. iommu_snooping is not initialized to zero when vm_iommu_init() called.
So it is possible to get a random value.
2. One line added by commit 2c2e2c38("IOMMU Identity Mapping Support")
change the code path, let it bypass domain_update_iommu_cap(), as well as
missing the increment of domain iommu reference count.
The latter is also likely to cause a leak of domains on repeated VMM
assignment and deassignment.
Signed-off-by: Sheng Yang <sheng@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
The physical address passed to domain_pfn_mapping() should be rounded
down to the start of the MM page, not the VT-d page.
This issue causes kernel panic on PAGE_SIZE>VTD_PAGE_SIZE platforms e.g. ia64
platforms.
Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
In domain_sg_mapping(), use aligned_nrpages() instead of hand-coded
rounding code for calculating the size of each sg elem. This means that
on IA64 we correctly round up to the MM page size, not just to the VT-d
page size.
Also remove the incorrect mm_to_dma_pfn() when intel_map_sg() calls
domain_sg_mapping() -- the 'size' variable is in VT-d pages already.
Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
This makes the hardware passthrough mode work a lot more like the
software version, so that the behaviour of a kernel with 'iommu=pt'
is the same whether the hardware supports passthrough or not.
In particular:
- We use a single si_domain for the pass-through devices.
- 32-bit devices can be taken out of the pass-through domain so that
they don't have to use swiotlb.
- Devices will work again after being removed from a KVM guest.
- A potential oops on OOM (in init_context_pass_through()) is fixed.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
The tboot module will DMA protect all of memory in order to ensure the that
kernel will be able to initialize without compromise (from DMA). Consequently,
the kernel must enable Intel Virtualization Technology for Directed I/O
(VT-d or Intel IOMMU) in order to replace this broad protection with the
appropriate page-granular protection. Otherwise DMA devices will be unable
to read or write from memory and the kernel will eventually panic.
Because runtime IOMMU support is configurable by command line options, this
patch will force it to be enabled regardless of the options specified, and will
log a message if it was required to force it on.
dmar.c | 7 +++++++
intel-iommu.c | 17 +++++++++++++++--
2 files changed, 22 insertions(+), 2 deletions(-)
Signed-off-by: Joseph Cihula <joseph.cihula@intel.com>
Signed-off-by: Shane Wang <shane.wang@intel.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
g_iommus is freed after we "goto error;".
Found by smatch (http://repo.or.cz/w/smatch.git).
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
We only ever obtain this lock immediately before the iova_rbtree_lock,
and release it immediately after the iova_rbtree_lock. So ditch it and
just use iova_rbtree_lock.
[v2: Remove the lockdep bits this time too]
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
After some API change, intel_iommu_unmap_range() introduced a assumption that
parameter size != 0, otherwise the dma_pte_clean_range() would have a
overflowed argument. But the user like KVM don't have this assumption before,
then some BUG() triggered.
Fix it by ignoring size = 0.
Signed-off-by: Sheng Yang <sheng@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We did before, in the end -- but it was at the bottom of a long stack of
functions. Add an inline wrapper get_valid_domain_for_dev() which will
use the cached one _first_ and only make the out-of-line call if it's
not already set.
This takes the average time taken for a 1-page intel_map_sg() from 5961
cycles to 4812 cycles on my Lenovo x200s test box -- a modest 20%.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Our current strategy for pass-through mode is to put all devices into
the 1:1 domain at startup (which is before we know what their dma_mask
will be), and only _later_ take them out of that domain, if it turns out
that they really can't address all of memory.
However, when there are a bunch of PCI devices behind a bridge, they all
end up with the same source-id on their DMA transactions, and hence in
the same IOMMU domain. This means that we _can't_ easily move them from
the 1:1 domain into their own domain at runtime, because there might be DMA
in-flight from their siblings.
So we have to adjust our pass-through strategy: For PCI devices not on
the root bus, and for the bridges which will take responsibility for
their transactions, we have to start up _out_ of the 1:1 domain, just in
case.
This fixes the BUG() we see when we have 32-bit-capable devices behind a
PCI-PCI bridge, and use the software identity mapping.
It does mean that we might end up using 'normal' mapping mode for some
devices which could actually live with the faster 1:1 mapping -- but
this is only for PCI devices behind bridges, which presumably aren't the
devices for which people are most concerned about performance.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
At boot time, the dma_mask won't have been set on any devices, so we
assume that all devices will be 64-bit capable (and thus get a 1:1 map).
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
This should fix kernel.org bug #11821, where the dcdbas driver makes up
a platform device and then uses dma_alloc_coherent() on it, in an
attempt to get memory < 4GiB.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
We need to give people a little more time to fix the broken drivers.
Re-introduce this, but tied in properly with the 'iommu=pt' support this
time. Change the config option name and make it default to 'no' too.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
We do this twice, and it's about to get more complicated. This makes the
code slightly clearer about what it's doing, too.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
When we reattach a device to the si_domain (because it's been removed
from a VM), we weren't calling domain_context_mapping() to actually tell
the hardware about that.
We should really put the call to domain_context_mapping() into
domain_add_dev_info() -- we never call the latter without also doing the
former, and we can keep the error paths simple that way. But that's a
cleanup which can wait for 2.6.32 now.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
We should check iommu_dummy() _first_, because that means it's attached
to an iommu that we've just disabled completely. At the moment, we might
try to put the device into the identity mapping domain.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
The aligned_nrpages() function rounds up to the next VM page, but
returns its result as a number of DMA pages.
Purely theoretical except on IA64, which doesn't boot with VT-d right
now anyway.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Check dma_pte_present() and only free the page if there _is_ one.
Kind of surprising that there was no warning about this.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
On Wed, 2009-07-01 at 16:59 -0700, Linus Torvalds wrote:
> I also _really_ hate how you do
>
> (unsigned long)pte >> VTD_PAGE_SHIFT ==
> (unsigned long)first_pte >> VTD_PAGE_SHIFT
Kill this, in favour of just looking to see if the incremented pte
pointer has 'wrapped' onto the next page. Which means we have to check
it _after_ incrementing it, not before.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
This would have found the bug in i386 pci_unmap_addr() a long time ago.
We shouldn't just silently return without doing anything.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Since we're using cmpxchg64() anyway (because that's the only way to do
an atomic 64-bit store on i386), we might as well ditch the extra
locking and just use cmpxchg64() to ensure that we don't add the page
twice.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
As with other functions, batch the CPU data cache flushes and don't keep
recalculating PTE addresses.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
The loop condition was wrong -- we should free a PMD only if its
_entire_ range is within the range we're intending to clear. The
early-termination condition was right, but not the loop.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Instead of calling domain_pfn_mapping() repeatedly with single or
small numbers of pages, just pass the sglist in. It can optimise the
number of cache flushes like domain_pfn_mapping() does, and gives a huge
speedup for large scatterlists.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
There's no need for the separate iommu_alloc_iova() function, and
certainly not for it to be global. Remove the underscores while we're at
it.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
As with dma_pte_clear_range(), don't keep flushing a single PTE at a
time. And also micro-optimise the setting of PTE values rather than
using the helper functions to do all the masking.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
It's a bit silly to repeatedly call domain_flush_cache() for each PTE
individually, as we clear it. Instead, batch them up and flush a whole
range at a time. We might as well refrain from recalculating the PTE
address from scratch each time round the loop too.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
This is fairly broken anyway -- it doesn't take hotplug into account.
We should probably be checking page_is_ram() instead.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Kenji Kaneshige