- 13 Jan, 2015 1 commit
-
-
Jan Beulich authored
Using the native code here can't work properly, as the hypervisor would normally have cleared the two reason bits by the time Dom0 gets to see the NMI (if passed to it at all). There's a shared info field for this, and there's an existing hook to use - just fit the two together. This is particularly relevant so that NMIs intended to be handled by APEI / GHES actually make it to the respective handler. Note that the hook can (and should) be used irrespective of whether being in Dom0, as accessing port 0x61 in a DomU would be even worse, while the shared info field would just hold zero all the time. Note further that hardware NMI handling for PVH doesn't currently work anyway due to missing code in the hypervisor (but it is expected to work the native rather than the PV way). Signed-off-by:
Jan Beulich <jbeulich@suse.com> Reviewed-by:
Boris Ostrovsky <boris.ostrovsky@oracle.com> Signed-off-by:
David Vrabel <david.vrabel@citrix.com>
-
- 12 Jan, 2015 4 commits
-
-
Juergen Gross authored
With the introduction of the linear mapped p2m list setting memory areas to "invalid" had to be delayed. When doing the invalidation make sure no zero sized areas are processed. Signed-off-by:
Juegren Gross <jgross@suse.com> Signed-off-by:
-
Juergen Gross authored
When converting a pfn to a physical address be sure to use 64 bit wide types or convert the physical address to a pfn if possible. Signed-off-by:
-
Juergen Gross authored
When allocating a new pmd for the linear mapped p2m list a check is done for not introducing another pmd when this just happened on another cpu. In this case the old pte pointer was returned which points to the p2m_missing or p2m_identity page. The correct value would be the pointer to the found new page. Signed-off-by:
-
Juergen Gross authored
In xen_rebuild_p2m_list() for large areas of invalid or identity mapped memory the pmd entries on 32 bit systems are initialized wrong. Correct this error. Suggested-by:
-
- 08 Jan, 2015 4 commits
-
-
Vitaly Kuznetsov authored
In case kasprintf() fails in xen_setup_timer() we assign name to the static string "<timer kasprintf failed>". We, however, don't check that fact before issuing kfree() in xen_teardown_timer(), kernel is supposed to crash with 'kernel BUG at mm/slub.c:3341!' Solve the issue by making name a fixed length string inside struct xen_clock_event_device. 16 bytes should be enough. Suggested-by:
Laszlo Ersek <lersek@redhat.com> Signed-off-by:
Vitaly Kuznetsov <vkuznets@redhat.com> Signed-off-by:
-
David Vrabel authored
If the non-RAM regions in the e820 memory map are larger than the size of the initial balloon, a BUG was triggered as the frames are remaped beyond the limit of the linear p2m. The frames are remapped into the initial balloon area (xen_extra_mem) but not enough of this is available. Ensure enough extra memory regions are added for these remapped frames. Signed-off-by:
-
David Vrabel authored
This accounting is just used to print a diagnostic message that isn't very useful. Signed-off-by:
-
Boris Ostrovsky authored
With recent changes in p2m we now have legitimate cases when p2m memory needs to be freed during early boot (i.e. before slab is initialized). Signed-off-by:
-
- 23 Dec, 2014 1 commit
-
-
Boris Ostrovsky authored
There is no reason for having it and, with commit 250a1ac6 ("x86, smpboot: Remove pointless preempt_disable() in native_smp_prepare_cpus()"), it prevents HVM guests from booting. Signed-off-by:
-
- 15 Dec, 2014 1 commit
-
-
David Vrabel authored
-
- 11 Dec, 2014 2 commits
-
-
Juergen Gross authored
With the virtual mapped linear p2m list the post-init mmu operations must be used for setting up the p2m mappings, as in case of CONFIG_FLATMEM the init routines may trigger BUGs. paging_init() sets up all infrastructure needed to switch to the post-init mmu ops done by xen_post_allocator_init(). With the virtual mapped linear p2m list we need some mmu ops during setup of this list, so we have to switch to the correct mmu ops as soon as possible. The p2m list is usable from the beginning, just expansion requires to have established the new linear mapping. So the call of xen_remap_memory() had to be introduced, but this is not due to the mmu ops requiring this. Summing it up: calling xen_post_allocator_init() not directly after paging_init() was conceptually wrong in the beginning, it just didn't matter up to now as no functions used between the two calls needed some critical mmu ops (e.g. alloc_pte). This has changed now, so I corrected it. Reported-by:
-
David Vrabel authored
This reverts commit 2c3fc8d2. This commit broke on x86 PV because entries in the generic SWIOTLB are indexed using (pseudo-)physical address not DMA address and these are not the same in a x86 PV guest. Signed-off-by:
Stefano Stabellini <stefano.stabellini@eu.citrix.com>
-
- 10 Dec, 2014 1 commit
-
-
David Vrabel authored
This reverts commit 2c3fc8d2. This commit broke on x86 PV because entries in the generic SWIOTLB are indexed using (pseudo-)physical address not DMA address and these are not the same in a x86 PV guest. Signed-off-by:
-
- 08 Dec, 2014 2 commits
-
-
Juergen Gross authored
Commit 5b8e7d80 removed the __init annotation from xen_set_identity_and_remap_chunk(). Add it again. Signed-off-by:
-
Juergen Gross authored
Introduce two helper functions to safely read and write unsigned long values from or to memory when the access may fault because the mapping is non-present or read-only. These helpers can be used instead of open coded uses of __get_user() and __put_user() avoiding the need to do casts to fix sparse warnings. Use the helpers in page.h and p2m.c. This will fix the sparse warnings when doing "make C=1". Signed-off-by:
-
- 04 Dec, 2014 24 commits
-
-
Juergen Gross authored
Instead of checking at each call of set_phys_to_machine() whether a new p2m page has to be allocated due to writing an entry in a large invalid or identity area, just map those areas read only and react to a page fault on write by allocating the new page. This change will make the common path with no allocation much faster as it only requires a single write of the new mfn instead of walking the address translation tables and checking for the special cases. Suggested-by:
Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Signed-off-by:
-
Juergen Gross authored
At start of the day the Xen hypervisor presents a contiguous mfn list to a pv-domain. In order to support sparse memory this mfn list is accessed via a three level p2m tree built early in the boot process. Whenever the system needs the mfn associated with a pfn this tree is used to find the mfn. Instead of using a software walked tree for accessing a specific mfn list entry this patch is creating a virtual address area for the entire possible mfn list including memory holes. The holes are covered by mapping a pre-defined page consisting only of "invalid mfn" entries. Access to a mfn entry is possible by just using the virtual base address of the mfn list and the pfn as index into that list. This speeds up the (hot) path of determining the mfn of a pfn. Kernel build on a Dell Latitude E6440 (2 cores, HT) in 64 bit Dom0 showed following improvements: Elapsed time: 32:50 -> 32:35 System: 18:07 -> 17:47 User: 104:00 -> 103:30 Tested with following configurations: - 64 bit dom0, 8GB RAM - 64 bit dom0, 128 GB RAM, PCI-area above 4 GB - 32 bit domU, 512 MB, 8 GB, 43 GB (more wouldn't work even without the patch) - 32 bit domU, ballooning up and down - 32 bit domU, save and restore - 32 bit domU with PCI passthrough - 64 bit domU, 8 GB, 2049 MB, 5000 MB - 64 bit domU, ballooning up and down - 64 bit domU, save and restore - 64 bit domU with PCI passthrough Signed-off-by: -
Juergen Gross authored
Today get_phys_to_machine() is always called when the mfn for a pfn is to be obtained. Add a wrapper __pfn_to_mfn() as inline function to be able to avoid calling get_phys_to_machine() when possible as soon as the switch to a linear mapped p2m list has been done. Signed-off-by:
-
Juergen Gross authored
Introduces lookup_pmd_address() to get the address of the pmd entry related to a virtual address in the current address space. This function is needed for support of a virtual mapped sparse p2m list in xen pv domains, as we need the address of the pmd entry, not the one of the pte in that case. Signed-off-by:
-
Juergen Gross authored
When the physical memory configuration is initialized the p2m entries for not pouplated memory pages are set to "invalid". As those pages are beyond the hypervisor built p2m list the p2m tree has to be extended. This patch delays processing the extra memory related p2m entries during the boot process until some more basic memory management functions are callable. This removes the need to create new p2m entries until virtual memory management is available. Signed-off-by:
-
Juergen Gross authored
The m2p overrides are used to be able to find the local pfn for a foreign mfn mapped into the domain. They are used by driver backends having to access frontend data. As this functionality isn't used in early boot it makes no sense to initialize the m2p override functions very early. It can be done later without doing any harm, removing the need for allocating memory via extend_brk(). While at it make some m2p override functions static as they are only used internally. Signed-off-by:
-
Juergen Gross authored
Early in the boot process the memory layout of a pv-domain is changed to match the E820 map (either the host one for Dom0 or the Xen one) regarding placement of RAM and PCI holes. This requires removing memory pages initially located at positions not suitable for RAM and adding them later at higher addresses where no restrictions apply. To be able to operate on the hypervisor supported p2m list until a virtual mapped linear p2m list can be constructed, remapping must be delayed until virtual memory management is initialized, as the initial p2m list can't be extended unlimited at physical memory initialization time due to it's fixed structure. A further advantage is the reduction in complexity and code volume as we don't have to be careful regarding memory restrictions during p2m updates. Signed-off-by:
-
Juergen Gross authored
In arch/x86/xen/p2m.c three different allocation functions for obtaining a memory page are used: extend_brk(), alloc_bootmem_align() or __get_free_page(). Which of those functions is used depends on the progress of the boot process of the system. Introduce a common allocation routine selecting the to be called allocation routine dynamically based on the boot progress. This allows moving initialization steps without having to care about changing allocation calls. Signed-off-by:
-
Juergen Gross authored
Some functions in arch/x86/xen/p2m.c are used locally only. Make them static. Rearrange the functions in p2m.c to avoid forward declarations. Signed-off-by:
-
Juergen Gross authored
The source arch/x86/xen/p2m.c has some coding style issues. Fix them. Signed-off-by:
-
Boris Ostrovsky authored
When hardware supports APIC/x2APIC virtualization we don't need to use pirqs for MSI handling and instead use APIC since most APIC accesses (MMIO or MSR) will now be processed without VMEXITs. As an example, netperf on the original code produces this profile (collected wih 'xentrace -e 0x0008ffff -T 5'): 342 cpu_change 260 CPUID 34638 HLT 64067 INJ_VIRQ 28374 INTR 82733 INTR_WINDOW 10 NPF 24337 TRAP 370610 vlapic_accept_pic_intr 307528 VMENTRY 307527 VMEXIT 140998 VMMCALL 127 wrap_buffer After applying this patch the same test shows 230 cpu_change 260 CPUID 36542 HLT 174 INJ_VIRQ 27250 INTR 222 INTR_WINDOW 20 NPF 24999 TRAP 381812 vlapic_accept_pic_intr 166480 VMENTRY 166479 VMEXIT 77208 VMMCALL 81 wrap_buffer ApacheBench results (ab -n 10000 -c 200) improve by about 10% Signed-off-by:Andrew Cooper <andrew.cooper3@citrix.com> Signed-off-by:
-
Boris Ostrovsky authored
If the hardware supports APIC virtualization we may decide not to use pirqs and instead use APIC/x2APIC directly, meaning that we don't want to set x86_msi.setup_msi_irqs and x86_msi.teardown_msi_irq to Xen-specific routines. However, x2APIC is not set up by the time pci_xen_hvm_init() is called so we need to postpone setting these ops until later, when we know which APIC mode is used. (Note that currently x2APIC is never initialized on HVM guests. This may change in the future) Signed-off-by:
-
Jan Beulich authored
When a PF driver unloads, it may find it necessary to leave the VFs around simply because of pciback having marked them as assigned to a guest. Utilize a suitable notification to let go of the VFs, thus allowing the PF to go back into the state it was before its driver loaded (which in particular allows the driver to be loaded again with it being able to create the VFs anew, but which also allows to then pass through the PF instead of the VFs). Don't do this however for any VFs currently in active use by a guest. Signed-off-by:
-
Konrad Rzeszutek Wilk authored
The commit "xen/pciback: Don't deadlock when unbinding." was using the version of pci_reset_function which would lock the device lock. That is no good as we can dead-lock. As such we swapped to using the lock-less version and requiring that the callers of 'pcistub_put_pci_dev' take the device lock. And as such this bug got exposed. Using the lock-less version is OK, except that we tried to use 'pci_restore_state' after the lock-less version of __pci_reset_function_locked - which won't work as 'state_saved' is set to false. Said 'state_saved' is a toggle boolean that is to be used by the sequence of a) pci_save_state/pci_restore_state or b) pci_load_and_free_saved_state/pci_restore_state. We don't want to use a) as the guest might have messed up the PCI configuration space and we want it to revert to the state when the PCI device was binded to us. Therefore we pick b) to restore the configuration space. We restore from our 'golden' version of PCI configuration space, when an: - Device is unbinded from pciback - Device is detached from a guest. Reported-by:
Sander Eikelenboom <linux@eikelenboom.it> Signed-off-by:
-
Konrad Rzeszutek Wilk authored
We have the pci_load_and_free_saved_state, and pci_store_saved_state but are missing the functionality to just load the state multiple times in the PCI device without having to free/save the state. This patch makes it possible to use this function. Signed-off-by:
Bjorn Helgaas <bhelgaas@google.com> Signed-off-by:
-
Konrad Rzeszutek Wilk authored
A little cleanup. No functional difference. Reviewed-by:
-
Konrad Rzeszutek Wilk authored
We had been printing it only if the device was built with debug enabled. But this information is useful in the field to troubleshoot. Signed-off-by:
-
Konrad Rzeszutek Wilk authored
Cleanup the function a bit - also include the id of the domain that is using the device. Signed-off-by:
-
Konrad Rzeszutek Wilk authored
Instead of open-coding it in drivers that want to double check that their functions are indeed holding the device lock. Signed-off-by:
Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by:
-
Konrad Rzeszutek Wilk authored
As commit 0a9fd015 'xen/pciback: Document the entry points for 'pcistub_put_pci_dev'' explained there are four entry points in this function. Two of them are when the user fiddles in the SysFS to unbind a device which might be in use by a guest or not. Both 'unbind' states will cause a deadlock as the the PCI lock has already been taken, which then pci_device_reset tries to take. We can simplify this by requiring that all callers of pcistub_put_pci_dev MUST hold the device lock. And then we can just call the lockless version of pci_device_reset. To make it even simpler we will modify xen_pcibk_release_pci_dev to quality whether it should take a lock or not - as it ends up calling xen_pcibk_release_pci_dev and needs to hold the lock. Reviewed-by:
-
Stefano Stabellini authored
Need to pass the pointer within the swiotlb internal buffer to the swiotlb library, that in the case of xen_unmap_single is dev_addr, not paddr. Signed-off-by:
-
Stefano Stabellini authored
In xen_swiotlb_sync_single we always call xen_dma_sync_single_for_cpu, even when we should call xen_dma_sync_single_for_device. Fix that. Signed-off-by:
-
Stefano Stabellini authored
On x86 truncation cannot occur because config XEN depends on X86_64 || (X86_32 && X86_PAE). On ARM truncation can occur without CONFIG_ARM_LPAE, when the dma operation involves foreign grants. However in that case the physical address returned by xen_bus_to_phys is actually invalid (there is no mfn to pfn tracking for foreign grants on ARM) and it is not used. Signed-off-by:
Catalin Marinas <catalin.marinas@arm.com> Acked-by:
-
Stefano Stabellini authored
xen_dma_unmap_page and xen_dma_sync_single_for_cpu take a dma_addr_t handle as argument, not a physical address. Signed-off-by:
-
