Dan Carpenter [Sat, 8 Sep 2012 09:57:35 +0000 (12:57 +0300)]
xen/privcmd: return -EFAULT on error
__copy_to_user() returns the number of bytes remaining to be copied but
we want to return a negative error code here.
Acked-by: Andres Lagar-Cavilla <andres@lagarcavilla.org> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
(cherry picked from commit 9d2be9287107695708e6aae5105a8a518a6cb4d0)
Andres Lagar-Cavilla [Thu, 6 Sep 2012 17:24:39 +0000 (13:24 -0400)]
xen/privcmd: Fix mmap batch ioctl error status copy back.
Copy back of per-slot error codes is only necessary for V2. V1 does not provide
an error array, so copyback will unconditionally set the global rc to EFAULT.
Only copyback for V2.
Signed-off-by: Andres Lagar-Cavilla <andres@lagarcavilla.org> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
(cherry picked from commit 1714df7f2cee6a741c3ed24231ec5db25b90633a)
Andres Lagar-Cavilla [Fri, 31 Aug 2012 13:59:30 +0000 (09:59 -0400)]
xen/privcmd: add PRIVCMD_MMAPBATCH_V2 ioctl
PRIVCMD_MMAPBATCH_V2 extends PRIVCMD_MMAPBATCH with an additional
field for reporting the error code for every frame that could not be
mapped. libxc prefers PRIVCMD_MMAPBATCH_V2 over PRIVCMD_MMAPBATCH.
Also expand PRIVCMD_MMAPBATCH to return appropriate error-encoding top nibble
in the mfn array.
Signed-off-by: David Vrabel <david.vrabel@citrix.com> Signed-off-by: Andres Lagar-Cavilla <andres@lagarcavilla.org> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
(cherry picked from commit ceb90fa0a8008059ecbbf9114cb89dc71a730bb6)
David Vrabel [Thu, 30 Aug 2012 12:58:11 +0000 (13:58 +0100)]
xen/mm: return more precise error from xen_remap_domain_range()
Callers of xen_remap_domain_range() need to know if the remap failed
because frame is currently paged out. So they can retry the remap
later on. Return -ENOENT in this case.
This assumes that the error codes returned by Xen are a subset of
those used by the kernel. It is unclear if this is defined as part of
the hypercall ABI.
Acked-by: Andres Lagar-Cavilla <andres@lagarcavilla.org> Signed-off-by: David Vrabel <david.vrabel@citrix.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
(cherry picked from commit 69870a847856a1ba81f655a8633fce5f5b614730)
Konrad Rzeszutek Wilk [Mon, 13 Aug 2012 15:00:08 +0000 (11:00 -0400)]
xen/swiotlb: Fix compile warnings when using plain integer instead of NULL pointer.
arch/x86/xen/pci-swiotlb-xen.c:96:1: warning: Using plain integer as NULL pointer
arch/x86/xen/pci-swiotlb-xen.c:96:1: warning: Using plain integer as NULL pointer
Konrad Rzeszutek Wilk [Mon, 13 Aug 2012 17:26:11 +0000 (13:26 -0400)]
xen/swiotlb: Remove functions not needed anymore.
Sparse warns us off:
drivers/xen/swiotlb-xen.c:506:1: warning: symbol 'xen_swiotlb_map_sg' was not declared. Should it be static?
drivers/xen/swiotlb-xen.c:534:1: warning: symbol 'xen_swiotlb_unmap_sg' was not declared. Should it be static?
and it looks like we do not need this function at all.
Stefano Stabellini [Wed, 22 Aug 2012 16:20:16 +0000 (17:20 +0100)]
xen: allow privcmd for HVM guests
This patch removes the "return -ENOSYS" for auto_translated_physmap
guests from privcmd_mmap, thus it allows ARM guests to issue privcmd
mmap calls. However privcmd mmap calls are still going to fail for HVM
and hybrid guests on x86 because the xen_remap_domain_mfn_range
implementation is currently PV only.
Changes in v2:
- better commit message;
- return -EINVAL from xen_remap_domain_mfn_range if
auto_translated_physmap.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> Acked-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
(cherry picked from commit 1a1d43318aeb74d679372c0b65029957be274529)
Daniel De Graaf [Thu, 16 Aug 2012 20:40:26 +0000 (16:40 -0400)]
xen/sysfs: Use XENVER_guest_handle to query UUID
This hypercall has been present since Xen 3.1, and is the preferred
method for a domain to obtain its UUID. Fall back to the xenstore method
if using an older version of Xen (which returns -ENOSYS).
Signed-off-by: Daniel De Graaf <dgdegra@tycho.nsa.gov> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
(cherry picked from commit 5c13f8067745efc15f6ad0158b58d57c44104c25)
Stefano Stabellini [Mon, 6 Aug 2012 14:27:24 +0000 (15:27 +0100)]
xen: update xen_add_to_physmap interface
Update struct xen_add_to_physmap to be in sync with Xen's version of the
structure.
The size field was introduced by:
changeset: 24164:707d27fe03e7
user: Jean Guyader <jean.guyader@eu.citrix.com>
date: Fri Nov 18 13:42:08 2011 +0000
summary: mm: New XENMEM space, XENMAPSPACE_gmfn_range
According to the comment:
"This new field .size is located in the 16 bits padding between .domid
and .space in struct xen_add_to_physmap to stay compatible with older
versions."
Changes in v2:
- remove erroneous comment in the commit message.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
(cherry picked from commit b58aaa4b0b3506c094308342d746f600468c63d9)
xen/p2m: Fix one by off error in checking the P2M tree directory.
We would traverse the full P2M top directory (from 0->MAX_DOMAIN_PAGES
inclusive) when trying to figure out whether we can re-use some of the
P2M middle leafs.
Which meant that if the kernel was compiled with MAX_DOMAIN_PAGES=512
we would try to use the 512th entry. Fortunately for us the p2m_top_index
has a check for this:
Konrad Rzeszutek Wilk [Thu, 16 Aug 2012 20:38:55 +0000 (16:38 -0400)]
xen/p2m: When revectoring deal with holes in the P2M array.
When we free the PFNs and then subsequently populate them back
during bootup:
Freeing 20000-20200 pfn range: 512 pages freed
1-1 mapping on 20000->20200
Freeing 40000-40200 pfn range: 512 pages freed
1-1 mapping on 40000->40200
Freeing bad80-badf4 pfn range: 116 pages freed
1-1 mapping on bad80->badf4
Freeing badf6-bae7f pfn range: 137 pages freed
1-1 mapping on badf6->bae7f
Freeing bb000-100000 pfn range: 282624 pages freed
1-1 mapping on bb000->100000
Released 283999 pages of unused memory
Set 283999 page(s) to 1-1 mapping
Populating 1acb8a-1f20e9 pfn range: 283999 pages added
We end up having the P2M array (that is the one that was
grafted on the P2M tree) filled with IDENTITY_FRAME or
INVALID_P2M_ENTRY) entries. The patch titled
"xen/p2m: Reuse existing P2M leafs if they are filled with 1:1 PFNs or INVALID."
recycles said slots and replaces the P2M tree leaf's with
&mfn_list[xx] with p2m_identity or p2m_missing.
And re-uses the P2M array sections for other P2M tree leaf's.
For the above mentioned bootup excerpt, the PFNs at
0x20000->0x20200 are going to be IDENTITY based:
P2M[0][256][0] -> P2M[0][257][0] get turned in IDENTITY_FRAME.
We can re-use that and replace P2M[0][256] to point to p2m_identity.
The "old" page (the grafted P2M array provided by Xen) that was at
P2M[0][256] gets put somewhere else. Specifically at P2M[6][358],
b/c when we populate back:
we fill P2M[6][358][0] (and P2M[6][358], P2M[6][359], ...) with
the new MFNs.
That is all OK, except when we revector we assume that the PFN
count would be the same in the grafted P2M array and in the
newly allocated. Since that is no longer the case, as we have
holes in the P2M that point to p2m_missing or p2m_identity we
have to take that into account.
[v2: Check for overflow]
[v3: Move within the __va check]
[v4: Fix the computation] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
(cherry picked from commit 3fc509fc0c590900568ef516a37101d88f3476f5)
Konrad Rzeszutek Wilk [Fri, 17 Aug 2012 13:35:31 +0000 (09:35 -0400)]
xen/mmu: If the revector fails, don't attempt to revector anything else.
If the P2M revectoring would fail, we would try to continue on by
cleaning the PMD for L1 (PTE) page-tables. The xen_cleanhighmap
is greedy and erases the PMD on both boundaries. Since the P2M
array can share the PMD, we would wipe out part of the __ka
that is still used in the P2M tree to point to P2M leafs.
This fixes it by bypassing the revectoring and continuing on.
If the revector fails, a nice WARN is printed so we can still
troubleshoot this.
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Konrad Rzeszutek Wilk [Thu, 16 Aug 2012 20:38:55 +0000 (16:38 -0400)]
xen/p2m: When revectoring deal with holes in the P2M array.
When we free the PFNs and then subsequently populate them back
during bootup:
Freeing 20000-20200 pfn range: 512 pages freed
1-1 mapping on 20000->20200
Freeing 40000-40200 pfn range: 512 pages freed
1-1 mapping on 40000->40200
Freeing bad80-badf4 pfn range: 116 pages freed
1-1 mapping on bad80->badf4
Freeing badf6-bae7f pfn range: 137 pages freed
1-1 mapping on badf6->bae7f
Freeing bb000-100000 pfn range: 282624 pages freed
1-1 mapping on bb000->100000
Released 283999 pages of unused memory
Set 283999 page(s) to 1-1 mapping
Populating 1acb8a-1f20e9 pfn range: 283999 pages added
We end up having the P2M array (that is the one that was
grafted on the P2M tree) filled with IDENTITY_FRAME or
INVALID_P2M_ENTRY) entries. The patch titled
"xen/p2m: Reuse existing P2M leafs if they are filled with 1:1 PFNs or INVALID."
recycles said slots and replaces the P2M tree leaf's with
&mfn_list[xx] with p2m_identity or p2m_missing.
And re-uses the P2M array sections for other P2M tree leaf's.
For the above mentioned bootup excerpt, the PFNs at
0x20000->0x20200 are going to be IDENTITY based:
P2M[0][256][0] -> P2M[0][257][0] get turned in IDENTITY_FRAME.
We can re-use that and replace P2M[0][256] to point to p2m_identity.
The "old" page (the grafted P2M array provided by Xen) that was at
P2M[0][256] gets put somewhere else. Specifically at P2M[6][358],
b/c when we populate back:
we fill P2M[6][358][0] (and P2M[6][358], P2M[6][359], ...) with
the new MFNs.
That is all OK, except when we revector we assume that the PFN
count would be the same in the grafted P2M array and in the
newly allocated. Since that is no longer the case, as we have
holes in the P2M that point to p2m_missing or p2m_identity we
have to take that into account.
[v2: Check for overflow] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Konrad Rzeszutek Wilk [Fri, 17 Aug 2012 13:27:35 +0000 (09:27 -0400)]
xen/p2m: Reuse existing P2M leafs if they are filled with 1:1 PFNs or INVALID.
If P2M leaf is completly packed with INVALID_P2M_ENTRY or with
1:1 PFNs (so IDENTITY_FRAME type PFNs), we can swap the P2M leaf
with either a p2m_missing or p2m_identity respectively. The old
page (which was created via extend_brk or was grafted on from the
mfn_list) can be re-used for setting new PFNs.
This also means we can remove git commit: 5bc6f9888db5739abfa0cae279b4b442e4db8049
xen/p2m: Reserve 8MB of _brk space for P2M leafs when populating back
which tried to fix this.
and make the amount that is required to be reserved much smaller.
CC: stable@vger.kernel.org # for 3.5 only. Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Konrad Rzeszutek Wilk [Tue, 14 Aug 2012 20:37:31 +0000 (16:37 -0400)]
xen/mmu: Release just the MFN list, not MFN list and part of pagetables.
We call memblock_reserve for [start of mfn list] -> [PMD aligned end
of mfn list] instead of <start of mfn list> -> <page aligned end of mfn list].
This has the disastrous effect that if at bootup the end of mfn_list is
not PMD aligned we end up returning to memblock parts of the region
past the mfn_list array. And those parts are the PTE tables with
the disastrous effect of seeing this at bootup:
Write protecting the kernel read-only data: 10240k
Freeing unused kernel memory: 1860k freed
Freeing unused kernel memory: 200k freed
(XEN) mm.c:2429:d0 Bad type (saw 1400000000000002 != exp 7000000000000000) for mfn 116a80 (pfn 14e26)
...
(XEN) mm.c:908:d0 Error getting mfn 116a83 (pfn 14e2a) from L1 entry 8000000116a83067 for l1e_owner=0, pg_owner=0
(XEN) mm.c:908:d0 Error getting mfn 4040 (pfn 5555555555555555) from L1 entry 0000000004040601 for l1e_owner=0, pg_owner=0
.. and so on.
This is not for upstream as it memblock_x86_reserve_range is not
used upstream anymore.
When I back-ported the patches:
xen/x86: Use memblock_reserve for sensitive areas.
xen/mmu: Recycle the Xen provided L4, L3, and L2 pages
I simply used sed s/memblock_reserve/memblock_x86_reserve_range/.
That was incorrect as the parameters are different - memblock_reserve
as second expects the size, while memblock_x86_reserve_range expects
the physical address. This patch fixes those bugs.
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
We then try to populate those pages back. In the P2M tree however
the space for those leafs must be reserved - as such we use extend_brk.
We reserve 8MB of _brk space, which means we can fit over 1048576 PFNs - which is more than we should ever need.
[v1: Made it 8MB of _brk space instead of 4MB per Jan's suggestion] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
(cherry picked from commit 99266871de5006ba7ad0bfece6bb283ede4094b9)
xen/mmu: Remove from __ka space PMD entries for pagetables.
Please first read the description in "xen/mmu: Copy and revector the
P2M tree."
At this stage, the __ka address space (which is what the old
P2M tree was using) is partially disassembled. The cleanup_highmap
has removed the PMD entries from 0-16MB and anything past _brk_end
up to the max_pfn_mapped (which is the end of the ramdisk).
The xen_remove_p2m_tree and code around has ripped out the __ka for
the old P2M array.
Here we continue on doing it to where the Xen page-tables were.
It is safe to do it, as the page-tables are addressed using __va.
For good measure we delete anything that is within MODULES_VADDR
and up to the end of the PMD.
At this point the __ka only contains PMD entries for the start
of the kernel up to __brk.
[v1: Per Stefano's suggestion wrapped the MODULES_VADDR in debug] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
(cherry picked from commit 4e928e1a48b6b76e0b8384160213a32d03197e4b)
Please first read the description in "xen/p2m: Add logic to revector a
P2M tree to use __va leafs" patch.
The 'xen_revector_p2m_tree()' function allocates a new P2M tree
copies the contents of the old one in it, and returns the new one.
At this stage, the __ka address space (which is what the old
P2M tree was using) is partially disassembled. The cleanup_highmap
has removed the PMD entries from 0-16MB and anything past _brk_end
up to the max_pfn_mapped (which is the end of the ramdisk).
We have revectored the P2M tree (and the one for save/restore as well)
to use new shiny __va address to new MFNs. The xen_start_info
has been taken care of already in 'xen_setup_kernel_pagetable()' and
xen_start_info->shared_info in 'xen_setup_shared_info()', so
we are free to roam and delete PMD entries - which is exactly what
we are going to do. We rip out the __ka for the old P2M array.
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Conflicts:
As can be seen, the ramdisk, P2M and pagetables are taking
a bit of __ka addresses space. Which is a problem since the
MODULES_VADDR starts at 0xffffffffa0000000 - and P2M sits
right in there! This results during bootup with the inability to
load modules, with this error:
Since the __va and __ka are 1:1 up to MODULES_VADDR and
cleanup_highmap rids __ka of the ramdisk mapping, what
we want to do is similar - get rid of the P2M in the __ka
address space. There are two ways of fixing this:
1) All P2M lookups instead of using the __ka address would
use the __va address. This means we can safely erase from
__ka space the PMD pointers that point to the PFNs for
P2M array and be OK.
2). Allocate a new array, copy the existing P2M into it,
revector the P2M tree to use that, and return the old
P2M to the memory allocate. This has the advantage that
it sets the stage for using XEN_ELF_NOTE_INIT_P2M
feature. That feature allows us to set the exact virtual
address space we want for the P2M - and allows us to
boot as initial domain on large machines.
So we pick option 2).
This patch only lays the groundwork in the P2M code. The patch
that modifies the MMU is called "xen/mmu: Copy and revector the P2M tree."
xen/mmu: Recycle the Xen provided L4, L3, and L2 pages
As we are not using them. We end up only using the L1 pagetables
and grafting those to our page-tables.
[v1: Per Stefano's suggestion squashed two commits]
[v2: Per Stefano's suggestion simplified loop] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Conflicts:
Acked-by: Jan Beulich <jbeulich@suse.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
(cherry picked from commit cbc09be35990fb3d15671507f11c3e90479ef816)
xen/mmu: Provide comments describing the _ka and _va aliasing issue
Which is that the level2_kernel_pgt (__ka virtual addresses)
and level2_ident_pgt (__va virtual address) contain the same
PMD entries. So if you modify a PTE in __ka, it will be reflected
in __va (and vice-versa).
xen/x86: Use memblock_reserve for sensitive areas.
instead of a big memblock_reserve. This way we can be more
selective in freeing regions (and it also makes it easier
to understand where is what).
[v1: Move the auto_translate_physmap to proper line]
[v2: Per Stefano suggestion add more comments] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
[upstream git commit 91addbf07abfdd109a9da4e02061e6ed3728b298]
Conflicts:
The P2M code is smart enough to return false (which means that it
cannot allocate anymore) and the error can perculate up the calling
stack without trouble - with the error logic doing the proper thing.
So check the __brk_limit values before allocating from extend_brk.
This allows us to boot on machines where we do not have enough
__brk space, and we would get this:
Interestingly enough, most of the time we are not going to hit this
b/c the _brk space is quite large (v3.5): ffffffff81a25000 B __brk_base ffffffff81e43000 B __brk_limit
= ~4MB.
vs earlier kernels (with this back-ported), the space is smaller: ffffffff81a25000 B __brk_base ffffffff81a7b000 B __brk_limit
= 344 kBytes.
With this patch, we would get now a limited amount of pages populated back:
Freeing 9f-100 pfn range: 97 pages freed
Freeing b7ee0-ecd9b pfn range: 216763 pages freed
Released 216860 pages of unused memory
Set 295297 page(s) to 1-1 mapping
Populating 100000-134f1c pfn range: 30720 pages added
[while it was instructed to populate 216860 pages back
on this particular machine]
Olaf Hering [Tue, 17 Jul 2012 15:43:35 +0000 (17:43 +0200)]
xen PVonHVM: move shared_info to MMIO before kexec
Currently kexec in a PVonHVM guest fails with a triple fault because the
new kernel overwrites the shared info page. The exact failure depends on
the size of the kernel image. This patch moves the pfn from RAM into
MMIO space before the kexec boot.
The pfn containing the shared_info is located somewhere in RAM. This
will cause trouble if the current kernel is doing a kexec boot into a
new kernel. The new kernel (and its startup code) can not know where the
pfn is, so it can not reserve the page. The hypervisor will continue to
update the pfn, and as a result memory corruption occours in the new
kernel.
One way to work around this issue is to allocate a page in the
xen-platform pci device's BAR memory range. But pci init is done very
late and the shared_info page is already in use very early to read the
pvclock. So moving the pfn from RAM to MMIO is racy because some code
paths on other vcpus could access the pfn during the small window when
the old pfn is moved to the new pfn. There is even a small window were
the old pfn is not backed by a mfn, and during that time all reads
return -1.
Because it is not known upfront where the MMIO region is located it can
not be used right from the start in xen_hvm_init_shared_info.
To minimise trouble the move of the pfn is done shortly before kexec.
This does not eliminate the race because all vcpus are still online when
the syscore_ops will be called. But hopefully there is no work pending
at this point in time. Also the syscore_op is run last which reduces the
risk further.
Signed-off-by: Olaf Hering <olaf@aepfle.de> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Olaf Hering [Tue, 17 Jul 2012 09:59:15 +0000 (11:59 +0200)]
xen: simplify init_hvm_pv_info
init_hvm_pv_info is called only in PVonHVM context, move it into ifdef.
init_hvm_pv_info does not fail, make it a void function.
remove arguments from init_hvm_pv_info because they are not used by the
caller.
Signed-off-by: Olaf Hering <olaf@aepfle.de> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Olaf Hering [Tue, 10 Jul 2012 13:31:39 +0000 (15:31 +0200)]
xen: enable platform-pci only in a Xen guest
While debugging kexec issues in a PVonHVM guest I modified
xen_hvm_platform() to return false to disable all PV drivers. This
caused a crash in platform_pci_init() because it expects certain data
structures to be initialized properly.
To avoid such a crash make sure the driver is initialized only if
running in a Xen guest.
Signed-off-by: Olaf Hering <olaf@aepfle.de> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Olaf Hering [Tue, 10 Jul 2012 12:50:03 +0000 (14:50 +0200)]
xen/pv-on-hvm kexec: shutdown watches from old kernel
Add xs_reset_watches function to shutdown watches from old kernel after
kexec boot. The old kernel does not unregister all watches in the
shutdown path. They are still active, the double registration can not
be detected by the new kernel. When the watches fire, unexpected events
will arrive and the xenwatch thread will crash (jumps to NULL). An
orderly reboot of a hvm guest will destroy the entire guest with all its
resources (including the watches) before it is rebuilt from scratch, so
the missing unregister is not an issue in that case.
With this change the xenstored is instructed to wipe all active watches
for the guest. However, a patch for xenstored is required so that it
accepts the XS_RESET_WATCHES request from a client (see changeset
23839:42a45baf037d in xen-unstable.hg). Without the patch for xenstored
the registration of watches will fail and some features of a PVonHVM
guest are not available. The guest is still able to boot, but repeated
kexec boots will fail.
Signed-off-by: Olaf Hering <olaf@aepfle.de> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
David Vrabel [Mon, 9 Jul 2012 10:39:06 +0000 (11:39 +0100)]
xen/mm: zero PTEs for non-present MFNs in the initial page table
When constructing the initial page tables, if the MFN for a usable PFN
is missing in the p2m then that frame is initially ballooned out. In
this case, zero the PTE (as in decrease_reservation() in
drivers/xen/balloon.c).
This is obviously safe instead of having an valid PTE with an MFN of
INVALID_P2M_ENTRY (~0).
Signed-off-by: David Vrabel <david.vrabel@citrix.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
David Vrabel [Mon, 9 Jul 2012 10:39:05 +0000 (11:39 +0100)]
xen/mm: do direct hypercall in xen_set_pte() if batching is unavailable
In xen_set_pte() if batching is unavailable (because the caller is in
an interrupt context such as handling a page fault) it would fall back
to using native_set_pte() and trapping and emulating the PTE write.
On 32-bit guests this requires two traps for each PTE write (one for
each dword of the PTE). Instead, do one mmu_update hypercall
directly.
During construction of the initial page tables, continue to use
native_set_pte() because most of the PTEs being set are in writable
and unpinned pages (see phys_pmd_init() in arch/x86/mm/init_64.c) and
using a hypercall for this is very expensive.
This significantly improves page fault performance in 32-bit PV
guests.
lmbench3 test Before After Improvement
----------------------------------------------
lat_pagefault 3.18 us 2.32 us 27%
lat_proc fork 356 us 313.3 us 11%
Signed-off-by: David Vrabel <david.vrabel@citrix.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
David Vrabel [Mon, 9 Jul 2012 10:39:08 +0000 (11:39 +0100)]
x86/xen: avoid updating TLS descriptors if they haven't changed
When switching tasks in a Xen PV guest, avoid updating the TLS
descriptors if they haven't changed. This improves the speed of
context switches by almost 10% as much of the time the descriptors are
the same or only one is different.
The descriptors written into the GDT by Xen are modified from the
values passed in the update_descriptor hypercall so we keep shadow
copies of the three TLS descriptors to compare against.
lmbench3 test Before After Improvement
--------------------------------------------
lat_ctx -s 32 24 7.19 6.52 9%
lat_pipe 12.56 11.66 7%
Signed-off-by: David Vrabel <david.vrabel@citrix.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
xen: populate correct number of pages when across mem boundary (v2)
When populate pages across a mem boundary at bootup, the page count
populated isn't correct. This is due to mem populated to non-mem
region and ignored.
Pfn range is also wrongly aligned when mem boundary isn't page aligned.
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@oracle.com>
[v2: If xen_do_chunk fail(populate), abort this chunk and any others]
Suggested by David, thanks.
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Liu, Jinsong [Fri, 15 Jun 2012 01:03:39 +0000 (09:03 +0800)]
xen/mce: add .poll method for mcelog device driver
If a driver leaves its poll method NULL, the device is assumed to
be both readable and writable without blocking.
This patch add .poll method to xen mcelog device driver, so that
when mcelog use system calls like ppoll or select, it would be
blocked when no data available, and avoid spinning at CPU.
Reported-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Signed-off-by: Liu, Jinsong <jinsong.liu@intel.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Liu, Jinsong [Tue, 12 Jun 2012 15:11:16 +0000 (23:11 +0800)]
xen/mce: schedule a workqueue to avoid sleep in atomic context
copy_to_user might sleep and print a stack trace if it is executed
in an atomic spinlock context. Like this:
(XEN) CMCI: send CMCI to DOM0 through virq
BUG: sleeping function called from invalid context at /home/konradinux/kernel.h:199
in_atomic(): 1, irqs_disabled(): 0, pid: 4581, name: mcelog
Pid: 4581, comm: mcelog Tainted: G O 3.5.0-rc1upstream-00003-g149000b-dirty #1
[<ffffffff8109ad9a>] __might_sleep+0xda/0x100
[<ffffffff81329b0b>] xen_mce_chrdev_read+0xab/0x140
[<ffffffff81148945>] vfs_read+0xc5/0x190
[<ffffffff81148b0c>] sys_read+0x4c/0x90
[<ffffffff815bd039>] system_call_fastpath+0x16
This patch schedule a workqueue for IRQ handler to poll the data,
and use mutex instead of spinlock, so copy_to_user sleep in atomic
context would not occur.
[upstream git commit b3856ae3554ef4cea39126f52216182ee83e8ec8] Reported-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Suggested-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Signed-off-by: Liu, Jinsong <jinsong.liu@intel.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Liu, Jinsong [Thu, 7 Jun 2012 11:58:50 +0000 (19:58 +0800)]
x86, MCE, AMD: Adjust initcall sequence for xen
there are 3 funcs which need to be _initcalled in a logic sequence:
1. xen_late_init_mcelog
2. mcheck_init_device
3. threshold_init_device
xen_late_init_mcelog must register xen_mce_chrdev_device before
native mce_chrdev_device registration if running under xen platform;
mcheck_init_device should be inited before threshold_init_device to
initialize mce_device, otherwise a a NULL ptr dereference will cause panic.
so we use following _initcalls
1. device_initcall(xen_late_init_mcelog);
2. device_initcall_sync(mcheck_init_device);
3. late_initcall(threshold_init_device);
when running under xen, the initcall order is 1,2,3;
on baremetal, we skip 1 and we do only 2 and 3.
Liu, Jinsong [Thu, 7 Jun 2012 11:56:51 +0000 (19:56 +0800)]
xen/mce: Add mcelog support for Xen platform
When MCA error occurs, it would be handled by Xen hypervisor first,
and then the error information would be sent to initial domain for logging.
This patch gets error information from Xen hypervisor and convert
Xen format error into Linux format mcelog. This logic is basically
self-contained, not touching other kernel components.
By using tools like mcelog tool users could read specific error information,
like what they did under native Linux.
To test follow directions outlined in Documentation/acpi/apei/einj.txt
Konrad Rzeszutek Wilk [Fri, 25 May 2012 20:11:09 +0000 (16:11 -0400)]
xen/blkback: Copy id field when doing BLKIF_DISCARD.
We weren't copying the id field so when we sent the response
back to the frontend (especially with a 64-bit host and 32-bit
guest), we ended up using a random value. This lead to the
frontend crashing as it would try to pass to __blk_end_request_all
a NULL 'struct request' (b/c it would use the 'id' to find the
proper 'struct request' in its shadow array) and end up crashing:
BUG: unable to handle kernel NULL pointer dereference at 000000e4
IP: [<c0646d4c>] __blk_end_request_all+0xc/0x40
.. snip..
EIP is at __blk_end_request_all+0xc/0x40
.. snip..
[<ed95db72>] blkif_interrupt+0x172/0x330 [xen_blkfront]
This fixes the bug by passing in the proper id for the response.
Konrad Rzeszutek Wilk [Tue, 29 May 2012 16:36:43 +0000 (12:36 -0400)]
xen/balloon: Subtract from xen_released_pages the count that is populated.
We did not take into account that xen_released_pages would be
used outside the initial E820 parsing code. As such we would
did not subtract from xen_released_pages the count of pages
that we had populated back (instead we just did a simple
extra_pages = released - populated).
The balloon driver uses xen_released_pages to set the initial
current_pages count. If this is wrong (too low) then when a new
(higher) target is set, the balloon driver will request too many pages
from Xen."
This fixes errors such as:
(XEN) memory.c:133:d0 Could not allocate order=0 extent: id=0 memflags=0 (51 of 512)
during bootup and
free_memory : 0
where the free_memory should be 128.
Acked-by: David Vrabel <david.vrabel@citrix.com>
[v1: Per David's review made the git commit better] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Konrad Rzeszutek Wilk [Wed, 23 May 2012 17:28:44 +0000 (13:28 -0400)]
xen/events: Add WARN_ON when quick lookup found invalid type.
All of the bind_XYZ_to_irq do a quick lookup to see if the
event exists. And if they that value is returned instead of
initialized. This patch adds an extra logic to check that the
type returned is the proper one and we can use it to find
drivers that are doing something naught.
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Konrad Rzeszutek Wilk [Wed, 23 May 2012 16:56:59 +0000 (12:56 -0400)]
xen/hvc: Check HVM_PARAM_CONSOLE_[EVTCHN|PFN] for correctness.
We need to make sure that those parameters are setup to be correct.
As such the value of 0 is deemed invalid and we find that we
bail out. The hypervisor sets by default all of them to be zero
and when the hypercall is done does a simple:
a.value = d->arch.hvm_domain.params[a.index];
Which means that if the Xen toolstack forgot to setup the proper
HVM_PARAM_CONSOLE_EVTCHN, we would get the default value of 0
and use that.
CC: stable@kernel.org
Fixes-Oracle-Bug: 14091238 Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Stefano Stabellini [Mon, 21 May 2012 15:54:10 +0000 (16:54 +0100)]
xen: do not map the same GSI twice in PVHVM guests.
PV on HVM guests map GSIs into event channels. At restore time the
event channels are resumed by restore_pirqs.
Device drivers might try to register the same GSI again through ACPI at
restore time, but the GSI has already been mapped and bound by
restore_pirqs. This patch detects these situations and avoids
mapping the same GSI multiple times.
Without this patch we get:
(XEN) irq.c:2235: dom4: pirq 23 or emuirq 28 already mapped
and waste a pirq.
Daniel De Graaf [Tue, 8 May 2012 13:46:57 +0000 (09:46 -0400)]
xenbus: Add support for xenbus backend in stub domain
Add an ioctl to the /dev/xen/xenbus_backend device allowing the xenbus
backend to be started after the kernel has booted. This allows xenstore
to run in a different domain from the dom0.
Signed-off-by: Daniel De Graaf <dgdegra@tycho.nsa.gov> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Konrad Rzeszutek Wilk [Mon, 21 May 2012 13:19:38 +0000 (09:19 -0400)]
xen/smp: unbind irqworkX when unplugging vCPUs.
The git commit 1ff2b0c303698e486f1e0886b4d9876200ef8ca5
"xen: implement IRQ_WORK_VECTOR handler" added the functionality
to have a per-cpu "irqworkX" for the IPI APIC functionality.
However it missed the unbind when a vCPU is unplugged resulting
in an orphaned per-cpu interrupt line for unplugged vCPU:
Lin Ming [Mon, 30 Apr 2012 16:16:27 +0000 (00:16 +0800)]
xen/apic: implement io apic read with hypercall
Implements xen_io_apic_read with hypercall, so it returns proper
IO-APIC information instead of fabricated one.
Fallback to return an emulated IO_APIC values if hypercall fails.
[v2: fallback to return an emulated IO_APIC values if hypercall fails] Signed-off-by: Lin Ming <mlin@ss.pku.edu.cn> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Konrad Rzeszutek Wilk [Tue, 20 Mar 2012 22:53:10 +0000 (18:53 -0400)]
xen/x86: Implement x86_apic_ops
Or rather just implement one different function as opposed
to the native one : the read function.
We synthesize the values.
[upstream git commit 31b3c9d723407b395564d1fff3624cc0083ae520] Acked-by: Suresh Siddha <suresh.b.siddha@intel.com>
[v1: Rebased on top of tip/x86/urgent]
[v2: Return 0xfd instead of 0xff in the default case] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Conflicts:
arch/x86/xen/Makefile
arch/x86/xen/enlighten.c
arch/x86/xen/xen-ops.h
[In the makefile, the vga.o didn't exist in 3.0 - it was added
in 3.1]
Konrad Rzeszutek Wilk [Wed, 28 Mar 2012 16:37:36 +0000 (12:37 -0400)]
x86/apic: Replace io_apic_ops with x86_io_apic_ops.
Which makes the code fit within the rest of the x86_ops functions.
Acked-by: Suresh Siddha <suresh.b.siddha@intel.com>
[v1: Changed x86_apic -> x86_ioapic per Yinghai Lu <yinghai@kernel.org> suggestion]
[v2: Rebased on tip/x86/urgent and redid to match Ingo's syntax style] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Jeremy Fitzhardinge [Thu, 22 Mar 2012 02:58:08 +0000 (22:58 -0400)]
x86/ioapic: Add io_apic_ops driver layer to allow interception
Xen dom0 needs to paravirtualize IO operations to the IO APIC,
so add a io_apic_ops for it to intercept. Do this as ops
structure because there's at least some chance that another
paravirtualized environment may want to intercept these.
[upstream git commit 136d249ef7dbf0fefa292082cc40be1ea864cbd6] Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Acked-by: Suresh Siddha <suresh.b.siddha@intel.com> Cc: jwboyer@redhat.com Cc: yinghai@kernel.org Link: http://lkml.kernel.org/r/1332385090-18056-2-git-send-email-konrad.wilk@oracle.com
[ Made all the affected code easier on the eyes ] Signed-off-by: Ingo Molnar <mingo@kernel.org>
Conflicts:
Lin Ming [Fri, 20 Apr 2012 16:11:05 +0000 (00:11 +0800)]
xen: implement IRQ_WORK_VECTOR handler
[upstream git commit 1ff2b0c303698e486f1e0886b4d9876200ef8ca5] Signed-off-by: Lin Ming <mlin@ss.pku.edu.cn> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Ben Guthro [Fri, 20 Apr 2012 16:11:04 +0000 (00:11 +0800)]
xen: implement apic ipi interface
Map native ipi vector to xen vector.
Implement apic ipi interface with xen_send_IPI_one.
[upstream git commit f447d56d36af18c5104ff29dcb1327c0c0ac3634] Tested-by: Steven Noonan <steven@uplinklabs.net> Signed-off-by: Ben Guthro <ben@guthro.net> Signed-off-by: Lin Ming <mlin@ss.pku.edu.cn> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Conflicts:
Jan Beulich [Thu, 5 Apr 2012 15:10:07 +0000 (16:10 +0100)]
xen/gnttab: add deferred freeing logic
Rather than just leaking pages that can't be freed at the point where
access permission for the backend domain gets revoked, put them on a
list and run a timer to (infrequently) retry freeing them. (This can
particularly happen when unloading a frontend driver when devices are
still present, and the backend still has them in non-closed state or
hasn't finished closing them yet.)
[upstream git commit 569ca5b3f94cd0b3295ec5943aa457cf4a4f6a3a] Signed-off-by: Jan Beulich <jbeulich@suse.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
xen: enter/exit lazy_mmu_mode around m2p_override calls
This patch is a significant performance improvement for the
m2p_override: about 6% using the gntdev device.
Each m2p_add/remove_override call issues a MULTI_grant_table_op and a
__flush_tlb_single if kmap_op != NULL. Batching all the calls together
is a great performance benefit because it means issuing one hypercall
total rather than two hypercall per page.
If paravirt_lazy_mode is set PARAVIRT_LAZY_MMU, all these calls are
going to be batched together, otherwise they are issued one at a time.
Adding arch_enter_lazy_mmu_mode/arch_leave_lazy_mmu_mode around the
m2p_add/remove_override calls forces paravirt_lazy_mode to
PARAVIRT_LAZY_MMU, therefore makes sure that they are always batched.
However it is not safe to call arch_enter_lazy_mmu_mode if we are in
interrupt context or if we are already in PARAVIRT_LAZY_MMU mode, so
check for both conditions before doing so.
Changes in v4:
- rebased on 3.4-rc4: all the m2p_override users call gnttab_unmap_refs
and gnttab_map_refs;
- check whether we are in interrupt context and the lazy_mode we are in
before calling arch_enter/leave_lazy_mmu_mode.
Changes in v3:
- do not call arch_enter/leave_lazy_mmu_mode in xen_blkbk_unmap, that
can be called in interrupt context.
David Vrabel [Thu, 3 May 2012 15:15:42 +0000 (16:15 +0100)]
xen/setup: update VA mapping when releasing memory during setup
In xen_memory_setup(), if a page that is being released has a VA
mapping this must also be updated. Otherwise, the page will be not
released completely -- it will still be referenced in Xen and won't be
freed util the mapping is removed and this prevents it from being
reallocated at a different PFN.
This was already being done for the ISA memory region in
xen_ident_map_ISA() but on many systems this was omitting a few pages
as many systems marked a few pages below the ISA memory region as
reserved in the e820 map.
This fixes errors such as:
(XEN) page_alloc.c:1148:d0 Over-allocation for domain 0: 2097153 > 2097152
(XEN) memory.c:133:d0 Could not allocate order=0 extent: id=0 memflags=0 (0 of 17)
[upstream git commit 83d51ab473dddde7df858015070ed22b84ebe9a9] Signed-off-by: David Vrabel <david.vrabel@citrix.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
The fun comes when a PV guest that is run with a machine E820 - that
can either be the initial domain or a PCI PV guest, where the E820
looks like the normal thing:
Those 283999 pages are subtracted from the nr_pages and are returned
to the hypervisor. The end result is that the initial domain
boots with 1GB less memory as the nr_pages has been subtracted by
the amount of pages residing within the PCI hole. It can balloon up
to that if desired using 'xl mem-set 0 8092', but the balloon driver
is not always compiled in for the initial domain.
This patch, implements the populate hypercall (XENMEM_populate_physmap)
which increases the the domain with the same amount of pages that
were released.
The other solution (that did not work) was to transplant the MFN in
the P2M tree - the ones that were going to be freed were put in
the E820_RAM regions past the nr_pages. But the modifications to the
M2P array (the other side of creating PTEs) were not carried away.
As the hypervisor is the only one capable of modifying that and the
only two hypercalls that would do this are: the update_va_mapping
(which won't work, as during initial bootup only PFNs up to nr_pages
are mapped in the guest) or via the populate hypercall.
The end result is that the kernel can now boot with the
nr_pages without having to subtract the 283999 pages.
On a 8GB machine, with various dom0_mem= parameters this is what we get:
no dom0_mem
-Memory: 6485264k/9435136k available (5817k kernel code, 1136060k absent, 1813812k reserved, 2899k data, 696k init)
+Memory: 7619036k/9435136k available (5817k kernel code, 1136060k absent, 680040k reserved, 2899k data, 696k init)
Konrad Rzeszutek Wilk [Fri, 30 Mar 2012 19:37:07 +0000 (15:37 -0400)]
xen/setup: Only print "Freeing XXX-YYY pfn range: Z pages freed" if Z > 0
Otherwise we can get these meaningless:
Freeing bad80-badf4 pfn range: 0 pages freed
We also can do this for the summary ones - no point of printing
"Set 0 page(s) to 1-1 mapping"
[upstream git commit ca1182387e57470460294ce1e39e2d5518809811] Acked-by: David Vrabel <david.vrabel@citrix.com>
[v1: Extended to the summary printks] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Konrad Rzeszutek Wilk [Fri, 30 Mar 2012 18:33:14 +0000 (14:33 -0400)]
xen/p2m: An early bootup variant of set_phys_to_machine
During early bootup we can't use alloc_page, so to allocate
leaf pages in the P2M we need to use extend_brk. For that
we are utilizing the early_alloc_p2m and early_alloc_p2m_middle
functions to do the job for us. This function follows the
same logic as set_phys_to_machine.
Konrad Rzeszutek Wilk [Fri, 30 Mar 2012 18:15:14 +0000 (14:15 -0400)]
xen/p2m: Allow alloc_p2m_middle to call reserve_brk depending on argument
For identity cases we want to call reserve_brk only on the boundary
conditions of the middle P2M (so P2M[x][y][0] = extend_brk). This is
to work around identify regions (PCI spaces, gaps in E820) which are not
aligned on 2MB regions.
However for the case were we want to allocate P2M middle leafs at the
early bootup stage, irregardless of this alignment check we need some
means of doing that. For that we provide the new argument.
Ian Campbell [Wed, 14 Dec 2011 12:16:08 +0000 (12:16 +0000)]
xen: only limit memory map to maximum reservation for domain 0.
d312ae878b6a "xen: use maximum reservation to limit amount of usable RAM"
clamped the total amount of RAM to the current maximum reservation. This is
correct for dom0 but is not correct for guest domains. In order to boot a guest
"pre-ballooned" (e.g. with memory=1G but maxmem=2G) in order to allow for
future memory expansion the guest must derive max_pfn from the e820 provided by
the toolstack and not the current maximum reservation (which can reflect only
the current maximum, not the guest lifetime max). The existing algorithm
already behaves this correctly if we do not artificially limit the maximum
number of pages for the guest case.
David Vrabel [Wed, 28 Sep 2011 16:46:36 +0000 (17:46 +0100)]
xen: release all pages within 1-1 p2m mappings
In xen_memory_setup() all reserved regions and gaps are set to an
identity (1-1) p2m mapping. If an available page has a PFN within one
of these 1-1 mappings it will become inaccessible (as it MFN is lost)
so release them before setting up the mapping.
This can make an additional 256 MiB or more of RAM available
(depending on the size of the reserved regions in the memory map) if
the initial pages overlap with reserved regions.
The 1:1 p2m mappings are also extended to cover partial pages. This
fixes an issue with (for example) systems with a BIOS that puts the
DMI tables in a reserved region that begins on a non-page boundary.
[upstream git commit f3f436e] Signed-off-by: David Vrabel <david.vrabel@citrix.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
David Vrabel [Thu, 29 Sep 2011 11:26:19 +0000 (12:26 +0100)]
xen: allow extra memory to be in multiple regions
Allow the extra memory (used by the balloon driver) to be in multiple
regions (typically two regions, one for low memory and one for high
memory). This allows the balloon driver to increase the number of
available low pages (if the initial number if pages is small).
As a side effect, the algorithm for building the e820 memory map is
simpler and more obviously correct as the map supplied by the
hypervisor is (almost) used as is (in particular, all reserved regions
and gaps are preserved). Only RAM regions are altered and RAM regions
above max_pfn + extra_pages are marked as unused (the region is split
in two if necessary).
[upstream git commit dc91c72] Signed-off-by: David Vrabel <david.vrabel@citrix.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Conflicts:
David Vrabel [Wed, 28 Sep 2011 16:46:34 +0000 (17:46 +0100)]
xen: allow balloon driver to use more than one memory region
Allow the xen balloon driver to populate its list of extra pages from
more than one region of memory. This will allow platforms to provide
(for example) a region of low memory and a region of high memory.
The maximum possible number of extra regions is 128 (== E820MAX) which
is quite large so xen_extra_mem is placed in __initdata. This is safe
as both xen_memory_setup() and balloon_init() are in __init.
The balloon regions themselves are not altered (i.e., there is still
only the one region).
[upstream git commit 8b5d44a] Signed-off-by: David Vrabel <david.vrabel@citrix.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Conflicts:
drivers/xen/balloon.c
[The "Add support for pv hugepages and support for huge balloon pages." has
a lot of these fixes in]
David Vrabel [Fri, 4 May 2012 13:29:46 +0000 (14:29 +0100)]
xen/pci: don't use PCI BIOS service for configuration space accesses
The accessing PCI configuration space with the PCI BIOS32 service does
not work in PV guests.
On systems without MMCONFIG or where the BIOS hasn't marked the
MMCONFIG region as reserved in the e820 map, the BIOS service is
probed (even though direct access is preferred) and this hangs.
CC: stable@kernel.org Acked-by: Jan Beulich <jbeulich@suse.com> Signed-off-by: David Vrabel <david.vrabel@citrix.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
[upstream git commit 76a8df7b49168509df02461f83fab117a4a86e08]
Conflicts:
which is due to the fact we are trying to access a PFN that is not
accessible to us. The reason (at least in this case) was that
PGD[256] is set to __HYPERVISOR_VIRT_START which setup (by the
hypervisor) to point to a read-only linear map of the MFN->PFN array.
During our parsing we would get the MFN (a valid one), try to look
it up in the MFN->PFN tree and find it invalid and return ~0 as PFN.
Then pte_mfn_to_pfn would happilly feed that in, attach the flags
and return it back to the caller. In this case the ptdump_show
bitshifts it and we get this invalid value.
Instead of doing all of that, we detect the ~0 case and just return
!_PAGE_PRESENT.
CC: stable@kernel.org Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Konrad Rzeszutek Wilk [Wed, 2 May 2012 19:04:51 +0000 (15:04 -0400)]
xen/apic: Return the APIC ID (and version) for CPU 0.
On x86_64 on AMD machines where the first APIC_ID is not zero, we get:
ACPI: LAPIC (acpi_id[0x01] lapic_id[0x10] enabled)
BIOS bug: APIC version is 0 for CPU 1/0x10, fixing up to 0x10
BIOS bug: APIC version mismatch, boot CPU: 0, CPU 1: version 10
which means that when the ACPI processor driver loads and
tries to parse the _Pxx states it fails to do as, as it
ends up calling acpi_get_cpuid which does this:
for_each_possible_cpu(i) {
if (cpu_physical_id(i) == apic_id)
return i;
}
And the bootup CPU, has not been found so it fails and returns -1
for the first CPU - which then subsequently in the loop that
"acpi_processor_get_info" does results in returning an error, which
means that "acpi_processor_add" failing and per_cpu(processor)
is never set (and is NULL).
That means that when xen-acpi-processor tries to load (much much
later on) and parse the P-states it gets -ENODEV from
acpi_processor_register_performance() (which tries to read
the per_cpu(processor)) and fails to parse the data.
Reported-by-and-Tested-by: Stefan Bader <stefan.bader@canonical.com> Suggested-by: Boris Ostrovsky <boris.ostrovsky@amd.com>
[v2: Bit-shift APIC ID by 24 bits] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
The reason we are dying is b/c the call acpi_get_override_irq() is used,
which returns the polarity and trigger for the IRQs. That function calls
mp_find_ioapics to get the 'struct ioapic' structure - which along with the
mp_irq[x] is used to figure out the default values and the polarity/trigger
overrides. Since the mp_find_ioapics now returns -1 [b/c the IOAPIC is filled
with 0xffffffff], the acpi_get_override_irq() stops trying to lookup in the
mp_irq[x] the proper INT_SRV_OVR and we can't install the SCI interrupt.
The proper fix for this is going in v3.5 and adds an x86_io_apic_ops
struct so that platforms can override it. But for v3.4 lets carry this
work-around. This patch does that by providing a slightly different variant
of the fake IOAPIC entries.
[upstream git commit] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
David Vrabel [Thu, 26 Apr 2012 18:44:06 +0000 (19:44 +0100)]
xen: correctly check for pending events when restoring irq flags
In xen_restore_fl_direct(), xen_force_evtchn_callback() was being
called even if no events were pending. This resulted in (depending on
workload) about a 100 times as many xen_version hypercalls as
necessary.
Fix this by correcting the sense of the conditional jump.
This seems to give a significant performance benefit for some
workloads.
There is some subtle tricksy "..since the check here is trying to
check both pending and masked in a single cmpw, but I think this is
correct. It will call check_events now only when the combined
mask+pending word is 0x0001 (aka unmasked, pending)." (Ian)
[upstream git commit 7eb7ce4d2e8991aff4ecb71a81949a907ca755ac] CC: stable@kernel.org Acked-by: Ian Campbell <ian.campbell@citrix.com> Signed-off-by: David Vrabel <david.vrabel@citrix.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
xen/smp: Fix crash when booting with ACPI hotplug CPUs.
When we boot on a machine that can hotplug CPUs and we
are using 'dom0_max_vcpus=X' on the Xen hypervisor line
to clip the amount of CPUs available to the initial domain,
we get this:
(XEN) Command line: com1=115200,8n1 dom0_mem=8G noreboot dom0_max_vcpus=8 sync_console mce_verbosity=verbose console=com1,vga loglvl=all guest_loglvl=all
.. snip..
DMI: Intel Corporation S2600CP/S2600CP, BIOS SE5C600.86B.99.99.x032.072520111118 07/25/2011
.. snip.
SMP: Allowing 64 CPUs, 32 hotplug CPUs
installing Xen timer for CPU 7
cpu 7 spinlock event irq 361
NMI watchdog: disabled (cpu7): hardware events not enabled
Brought up 8 CPUs
.. snip..
[acpi processor finds the CPUs are not initialized and starts calling
arch_register_cpu, which creates /sys/devices/system/cpu/cpu8/online]
CPU 8 got hotplugged
CPU 9 got hotplugged
CPU 10 got hotplugged
.. snip..
initcall 1_acpi_battery_init_async+0x0/0x1b returned 0 after 406 usecs
calling erst_init+0x0/0x2bb @ 1
[and the scheduler sticks newly started tasks on the new CPUs, but
said CPUs cannot be initialized b/c the hypervisor has limited the
amount of vCPUS to 8 - as per the dom0_max_vcpus=8 flag.
The spinlock tries to kick the other CPU, but the structure for that
is not initialized and we crash.]
BUG: unable to handle kernel paging request at fffffffffffffed8
IP: [<ffffffff81035289>] xen_spin_lock+0x29/0x60
PGD 180d067 PUD 180e067 PMD 0
Oops: 0002 [#1] SMP
CPU 7
Modules linked in:
xen: use the pirq number to check the pirq_eoi_map
In pirq_check_eoi_map use the pirq number rather than the Linux irq
number to check whether an eoi is needed in the pirq_eoi_map.
The reason is that the irq number is not always identical to the
pirq number so if we wrongly use the irq number to check the
pirq_eoi_map we are going to check for the wrong pirq to EOI.
As a consequence some interrupts might not be EOI'ed by the
guest correctly.
[upstream git commit 521394e4e679996955bc351cb6b64639751db2ff] Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> Tested-by: Tobias Geiger <tobias.geiger@vido.info>
[v1: Added some extra wording to git commit] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
drivers/block/xen-blkback/xenbus.c: In function 'xen_blkbk_discard':
drivers/block/xen-blkback/xenbus.c:419:4: warning: passing argument 1 of 'dev_warn' makes pointer from integer without a cast
+[enabled by default]
include/linux/device.h:894:5: note: expected 'const struct device *' but argument is of type 'long int'
It is unclear how that mistake made it in. It surely is wrong.
[upstream git commit a71e23d] Acked-by: Jens Axboe <axboe@kernel.dk> Reported-by: Stephen Rothwell <sfr@canb.auug.org.au> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Conflicts:
Konrad Rzeszutek Wilk [Wed, 14 Mar 2012 17:04:00 +0000 (13:04 -0400)]
xen/blkback: Make optional features be really optional.
They were using the xenbus_dev_fatal() function which would
change the state of the connection immediately. Which is not
what we want when we advertise optional features.
So make 'feature-discard','feature-barrier','feature-flush-cache'
optional.
[upstream git commit 3389bb8] Suggested-by: Jan Beulich <JBeulich@suse.com>
[v1: Made the discard function void and static] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Conflicts:
Jan Beulich [Thu, 5 Apr 2012 15:04:52 +0000 (16:04 +0100)]
xen-blkfront: module exit handling adjustments
The blkdev major must be released upon exit, or else the module can't
attach to devices using the same majors upon being loaded again. Also
avoid leaking the minor tracking bitmap.
[upstream git commit 4e55b3c] Signed-off-by: Jan Beulich <jbeulich@suse.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Jan Beulich [Thu, 5 Apr 2012 15:37:22 +0000 (16:37 +0100)]
xen-blkfront: properly name all devices
- devices beyond xvdzz didn't get proper names assigned at all
- extended devices with minors not representable within the kernel's
major/minor bit split spilled into foreign majors
[upstream git commit 85b6984] Signed-off-by: Jan Beulich <jbeulich@suse.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
xen-blkfront: set pages are FOREIGN_FRAME when sharing them
Set pages as FOREIGN_FRAME whenever blkfront shares them with another
domain. Then when blkfront un-share them, also removes the
FOREIGN_FRAME_BIT from the p2m.
We do it so that when the source and the destination domain are the same
(blkfront connected to a disk backend in the same domain) we can more easily
recognize which ones are the source pfns and which ones are the
destination pfns (both are going to be pointing to the same mfns).
Without this patch enstablishing a connection between blkfront and QEMU
qdisk in the same domain causes QEMU to hang and never return.
The scenario where this used is when a disk image in QCOW2 is used
for extracting the kernel and initrd image. The QCOW2 image file cannot
be loopback-ed and to run 'pygrub', the weird scaffolding of:
- setup QEMU and qdisk with the qcow2 image [disk backend]
- setup xen-blkfront mounting said disk backend in the domain.
- extract kernel and initrd
- tear it down.
The MFNs shared shared by the frontend are going to back two
different sets of PFNs: the original PFNs allocated by the frontend and
the new ones allocated by gntdev for the backend.
The problem is that when Linux calls mfn_to_pfn, passing as argument
one of the MFN shared by the frontend, we want to get the PFN returned by
m2p_find_override_pfn (that is the PFN setup by gntdev) but actually we
get the original PFN allocated by the frontend because considering that
the frontend and the backend are in the same domain:
One possible solution would be to always call m2p_find_override_pfn to
check out whether we have an entry for a given MFN. However it is not
very efficient or scalable.
The other option (that this patch is implementing) is to mark the pages
shared by the frontend as "foreign", so that mfn != mfn2.
It makes sense because from the frontend point of view they are donated
to the backend and while so they are not supposed to be used by the
frontend. In a way, they don't belong to the frontend anymore, at least
temporarily.
[upstream git commit 6a2c6177]
[v3: only set_phys_to_machine if xen_pv_domain] Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
[v1: Redid description a bit] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Steven Noonan [Fri, 17 Feb 2012 20:04:44 +0000 (12:04 -0800)]
xen-blkfront: make blkif_io_lock spinlock per-device
This patch moves the global blkif_io_lock to the per-device structure. The
spinlock seems to exists for two reasons: to disable IRQs when in the interrupt
handlers for blkfront, and to protect the blkfront VBDs when a detachment is
requested.
Having a global blkif_io_lock doesn't make sense given the use case, and it
drastically hinders performance due to contention. All VBDs with pending IOs
have to take the lock in order to get work done, which serializes everything
pretty badly.
[upstream git commit 3467811] Signed-off-by: Steven Noonan <snoonan@amazon.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
projects / users / jedix / linux-maple.git / log