- 03 May, 2016 1 commit
-
-
James Morse authored
If memory is located above 1<<VA_BITS, kvm adds an extra level to its page tables, merging the runtime tables and boot tables that contain the idmap. This lets us avoid the trampoline dance during initialisation. This also means there is no trampoline page mapped, so __cpu_reset_hyp_mode() can't call __kvm_hyp_reset() in this page. The good news is the idmap is still mapped, so we don't need the trampoline page. The bad news is we can't call it directly as the idmap is above HYP_PAGE_OFFSET, so its address is masked by kvm_call_hyp. Add a function __extended_idmap_trampoline which will branch into __kvm_hyp_reset in the idmap, change kvm_hyp_reset_entry() to return this address if __kvm_cpu_uses_extended_idmap(). In this case __kvm_hyp_reset() will still switch to the boot tables (which are the merged tables that were already in use), and branch into the idmap (where it already was). This fixes boot failures on these systems, where we fail to execute the missing trampoline page when tearing down kvm in init_subsystems(): [ 2.508922] kvm [1]: 8-bit VMID [ 2.512057] kvm [1]: Hyp mode initialized successfully [ 2.517242] kvm [1]: interrupt-controller@e1140000 IRQ13 [ 2.522622] kvm [1]: timer IRQ3 [ 2.525783] Kernel panic - not syncing: HYP panic: [ 2.525783] PS:200003c9 PC:0000007ffffff820 ESR:86000005 [ 2.525783] FAR:0000007ffffff820 HPFAR:00000000003ffff0 PAR:0000000000000000 [ 2.525783] VCPU: (null) [ 2.525783] [ 2.547667] CPU: 0 PID: 0 Comm: swapper/0 Tainted: G W 4.6.0-rc5+ #1 [ 2.555137] Hardware name: Default string Default string/Default string, BIOS ROD0084E 09/03/2015 [ 2.563994] Call trace: [ 2.566432] [<ffffff80080888d0>] dump_backtrace+0x0/0x240 [ 2.571818] [<ffffff8008088b24>] show_stack+0x14/0x20 [ 2.576858] [<ffffff80083423ac>] dump_stack+0x94/0xb8 [ 2.581899] [<ffffff8008152130>] panic+0x10c/0x250 [ 2.586677] [<ffffff8008152024>] panic+0x0/0x250 [ 2.591281] SMP: stopping secondary CPUs [ 3.649692] SMP: failed to stop secondary CPUs 0-2,4-7 [ 3.654818] Kernel Offset: disabled [ 3.658293] Memory Limit: none [ 3.661337] ---[ end Kernel panic - not syncing: HYP panic: [ 3.661337] PS:200003c9 PC:0000007ffffff820 ESR:86000005 [ 3.661337] FAR:0000007ffffff820 HPFAR:00000000003ffff0 PAR:0000000000000000 [ 3.661337] VCPU: (null) [ 3.661337] Reported-by:
Will Deacon <will.deacon@arm.com> Reviewed-by:
Marc Zyngier <marc.zyngier@arm.com> Signed-off-by:
James Morse <james.morse@arm.com> Signed-off-by:
-
- 28 Apr, 2016 17 commits
-
-
Ard Biesheuvel authored
Currently, our KASLR implementation randomizes the placement of the core kernel at 2 MB granularity. This is based on the arm64 kernel boot protocol, which mandates that the kernel is loaded TEXT_OFFSET bytes above a 2 MB aligned base address. This requirement is a result of the fact that the block size used by the early mapping code may be 2 MB at the most (for a 4 KB granule kernel) But we can do better than that: since a KASLR kernel needs to be relocated in any case, we can tolerate a physical misalignment as long as the virtual misalignment relative to this 2 MB block size is equal in size, and code to deal with this is already in place. Since we align the kernel segments to 64 KB, let's randomize the physical offset at 64 KB granularity as well (unless CONFIG_DEBUG_ALIGN_RODATA is enabled). This way, the page table and TLB footprint is not affected. The higher granularity allows for 5 bits of additional entropy to be used. Reviewed-by:
Matt Fleming <matt@codeblueprint.co.uk> Signed-off-by:
Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by:
-
Ezequiel Garcia authored
The rtc-lib dependency is not required, and seems it was just copy-pasted from ARM's Kconfig. If platform requires rtc-lib, they should select it individually. Reviewed-by:
Arnd Bergmann <arnd@arndb.de> Signed-off-by:
Ezequiel Garcia <ezequiel@vanguardiasur.com.ar> Signed-off-by:
-
Will Deacon authored
Selecting both DEBUG_PAGEALLOC and HIBERNATION results in a build failure: | kernel/built-in.o: In function `saveable_page': | memremap.c:(.text+0x100f90): undefined reference to `kernel_page_present' | kernel/built-in.o: In function `swsusp_save': | memremap.c:(.text+0x1026f0): undefined reference to `kernel_page_present' | make: *** [vmlinux] Error 1 James sayeth: "This is caused by DEBUG_PAGEALLOC, which clears the PTE_VALID bit from 'free' pages. Hibernate uses it as a hint that it shouldn't save/access that page. This function is used to test whether the PTE_VALID bit has been cleared by kernel_map_pages(), hibernate is the only user. Fixing this exposes a bigger problem with that configuration though: if the resume kernel has cut free pages out of the linear map, we copy this swiss-cheese view of memory, and try to use it to restore... We can fixup the copy of the linear map, but it then explodes in my lazy 'clean the whole kernel to PoC' after resume, as now both the kernel and linear map have holes in them." On closer inspection, the whole Kconfig machinery around DEBUG_PAGEALLOC, HIBERNATION, ARCH_SUPPORTS_DEBUG_PAGEALLOC and PAGE_POISONING looks like it might need some affection. In particular, DEBUG_ALLOC has: > depends on !HIBERNATION || ARCH_SUPPORTS_DEBUG_PAGEALLOC && !PPC && !SPARC which looks pretty fishy. For the moment, require ARCH_SUPPORTS_DEBUG_PAGEALLOC to depend on !HIBERNATION on arm64 and get allmodconfig building again. Signed-off-by:
-
James Morse authored
Hibernation represents a system state save/restore through a system reboot; this implies that the logical cpus carrying out hibernation/thawing must be the same, so that the context saved in the snapshot image on hibernation is consistent with the state of the system on resume. If resume from hibernation is driven through kernel command line parameter, the cpu responsible for thawing the system will be whatever CPU firmware boots the system on upon cold-boot (ie logical cpu 0); this means that in order to keep system context consistent between the hibernate snapshot image and system state on kernel resume from hibernate, logical cpu 0 must be online on hibernation and must be the logical cpu that creates the snapshot image. This patch adds a PM notifier that enforces logical cpu 0 is online when the hibernation is started (and prevents hibernation if it is not), which is sufficient to guarantee it will be the one creating the snapshot image therefore providing the resume cpu a consistent snapshot of the system to resume to. Signed-off-by:
Lorenzo Pieralisi <lorenzo.pieralisi@arm.com> Acked-by:
Catalin Marinas <catalin.marinas@arm.com> Signed-off-by:
-
James Morse authored
Add support for hibernate/suspend-to-disk. Suspend borrows code from cpu_suspend() to write cpu state onto the stack, before calling swsusp_save() to save the memory image. Restore creates a set of temporary page tables, covering only the linear map, copies the restore code to a 'safe' page, then uses the copy to restore the memory image. The copied code executes in the lower half of the address space, and once complete, restores the original kernel's page tables. It then calls into cpu_resume(), and follows the normal cpu_suspend() path back into the suspend code. To restore a kernel using KASLR, the address of the page tables, and cpu_resume() are stored in the hibernate arch-header and the el2 vectors are pivotted via the 'safe' page in low memory. Reviewed-by:
-
James Morse authored
Some architectures require code written to memory as if it were data to be 'cleaned' from any data caches before the processor can fetch them as new instructions. During resume from hibernate, the snapshot code copies some pages directly, meaning these architectures do not get a chance to perform their cache maintenance. Modify the read and decompress code to call flush_icache_range() on all pages that are restored, so that the restored in-place pages are guaranteed to be executable on these architectures. Signed-off-by:
Pavel Machek <pavel@ucw.cz> Acked-by:
Rafael J. Wysocki <rjw@rjwysocki.net> Acked-by:
-
Geoff Levand authored
Kexec and hibernate need to copy pages of memory, but may not have all of the kernel mapped, and are unable to call copy_page(). Add a simplistic copy_page() macro, that can be inlined in these situations. lib/copy_page.S provides a bigger better version, but uses more registers. Signed-off-by:
Geoff Levand <geoff@infradead.org> [Changed asm label to 9998, added commit message] Signed-off-by:
-
James Morse authored
KERNEL_START and KERNEL_END are useful outside head.S, move them to a header file. Signed-off-by:
-
James Morse authored
page.h uses '_AC' in the definition of PAGE_SIZE, but doesn't include linux/const.h where this is defined. This produces build warnings when only asm/page.h is included by asm code. Signed-off-by:
Mark Rutland <mark.rutland@arm.com> Acked-by:
-
James Morse authored
By enabling the MMU early in cpu_resume(), the sleep_save_sp and stack can be accessed by VA, which avoids the need to convert-addresses and clean to PoC on the suspend path. MMU setup is shared with the boot path, meaning the swapper_pg_dir is restored directly: ttbr1_el1 is no longer saved/restored. struct sleep_save_sp is removed, replacing it with a single array of pointers. cpu_do_{suspend,resume} could be further reduced to not restore: cpacr_el1, mdscr_el1, tcr_el1, vbar_el1 and sctlr_el1, all of which are set by __cpu_setup(). However these values all contain res0 bits that may be used to enable future features. Signed-off-by: -
James Morse authored
Hibernate could make use of the cpu_suspend() code to save/restore cpu state, however it needs to be able to return '0' from the 'finisher'. Rework cpu_suspend() so that the finisher is called from C code, independently from the save/restore of cpu state. Space to save the context in is allocated in the caller's stack frame, and passed into __cpu_suspend_enter(). Hibernate's use of this API will look like a copy of the cpu_suspend() function. Signed-off-by:
-
AKASHI Takahiro authored
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents kexec from rebooting the system at EL2. This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function, kvm_arch_hardware_disable() and kvm_arch_hardware_enable() respectively. We don't need the arm64 specific cpu hotplug hook any more. Since this patch modifies common code between arm and arm64, one stub definition, __cpu_reset_hyp_mode(), is added on arm side to avoid compilation errors. Signed-off-by:
AKASHI Takahiro <takahiro.akashi@linaro.org> [Rebase, added separate VHE init/exit path, changed resets use of kvm_call_hyp() to the __version, en/disabled hardware in init_subsystems(), added icache maintenance to __kvm_hyp_reset() and removed lr restore, removed guest-enter after teardown handling] Signed-off-by:
-
James Morse authored
A later patch implements kvm_arch_hardware_disable(), to remove kvm from el2, and re-instate the hyp-stub. This can happen while guests are running, particularly when kvm_reboot() calls kvm_arch_hardware_disable() on each cpu. This can interrupt a guest, remove kvm, then allow the guest to be scheduled again. This causes kvm_call_hyp() to be run against the hyp-stub. Change the hyp-stub to return a new exception type when this happens, and add code to kvm's handle_exit() to tell userspace we failed to enter the guest. Signed-off-by:
-
Geoff Levand authored
The existing arm64 hcall implementations are limited in that they only allow for two distinct hcalls; with the x0 register either zero or not zero. Also, the API of the hyp-stub exception vector routines and the KVM exception vector routines differ; hyp-stub uses a non-zero value in x0 to implement __hyp_set_vectors, whereas KVM uses it to implement kvm_call_hyp. To allow for additional hcalls to be defined and to make the arm64 hcall API more consistent across exception vector routines, change the hcall implementations to reserve all x0 values below 0xfff for hcalls such as {s,g}et_vectors(). Define two new preprocessor macros HVC_GET_VECTORS, and HVC_SET_VECTORS to be used as hcall type specifiers and convert the existing __hyp_get_vectors() and __hyp_set_vectors() routines to use these new macros when executing an HVC call. Also, change the corresponding hyp-stub and KVM el1_sync exception vector routines to use these new macros. Signed-off-by: -
James Morse authored
Today the 'hvc' calling KVM or the hyp-stub is expected to preserve all registers. KVM saves/restores the registers it needs on the EL2 stack using do_el2_call(). The hyp-stub has no stack, later patches need to be able to be able to clobber the link register. Move the link register save/restore to the the call sites. Signed-off-by:
-
Geoff Levand authored
We currently have macros defining flags for the arm64 sctlr registers in both kvm_arm.h and sysreg.h. To clean things up and simplify move the definitions of the SCTLR_EL2 flags from kvm_arm.h to sysreg.h, rename any SCTLR_EL1 or SCTLR_EL2 flags that are common to both registers to be SCTLR_ELx, with 'x' indicating a common flag, and fixup all files to include the proper header or to use the new macro names. Signed-off-by:
-
Geoff Levand authored
To allow the assembler macros defined in arch/arm64/mm/proc-macros.S to be used outside the mm code move the contents of proc-macros.S to asm/assembler.h. Also, delete proc-macros.S, and fix up all references to proc-macros.S. Signed-off-by:
-
- 26 Apr, 2016 8 commits
-
-
Ard Biesheuvel authored
If both ACPI and DT platform descriptions are available, and the kernel was configured at build time to support both flavours, the default policy is to prefer DT over ACPI, and preferring ACPI over DT while still allowing DT as a fallback is not possible. Since some enterprise features (such as RAS) depend on ACPI, it may be desirable for, e.g., distro installers to prefer ACPI boot but fall back to DT rather than failing completely if no ACPI tables are available. So introduce the 'acpi=on' kernel command line parameter for arm64, which signifies that ACPI should be used if available, and DT should only be used as a fallback. Acked-by:
-
Ard Biesheuvel authored
When booting a relocatable kernel image, there is no practical reason to refuse an image whose load address is not exactly TEXT_OFFSET bytes above a 2 MB aligned base address, as long as the physical and virtual misalignment with respect to the swapper block size are equal, and are both aligned to THREAD_SIZE. Since the virtual misalignment is under our control when we first enter the kernel proper, we can simply choose its value to be equal to the physical misalignment. So treat the misalignment of the physical load address as the initial KASLR offset, and fix up the remaining code to deal with that. Acked-by:
-
Ard Biesheuvel authored
For historical reasons, the kernel Image must be loaded into physical memory at a 512 KB offset above a 2 MB aligned base address. The region between the base address and the start of the kernel Image has no significance to the kernel itself, but it is currently mapped explicitly into the early kernel VMA range for all translation granules. In some cases (i.e., 4 KB granule), this is unavoidable, due to the 2 MB granularity of the early kernel mappings. However, in other cases, e.g., when running with larger page sizes, or in the future, with more granular KASLR, there is no reason to map it explicitly like we do currently. So update the logic so that the region is mapped only if that happens as a side effect of rounding the start address of the kernel to swapper block size, and leave it unmapped otherwise. Since the symbol kernel_img_size now simply resolves to the memory footprint of the kernel Image, we can drop its definition from image.h and opencode its calculation. Acked-by:
-
Ard Biesheuvel authored
When building a relocatable kernel, we currently rely on the fact that early 64-bit literal loads need to be deferred to after the relocation has been performed only if they involve symbol references, and not if they involve assemble time constants. While this is not an unreasonable assumption to make, it is better to switch to movk/movz sequences, since these are guaranteed to be resolved at link time, simply because there are no dynamic relocation types to describe them. Acked-by:
-
Ard Biesheuvel authored
Implement a macro mov_q that can be used to move an immediate constant into a 64-bit register, using between 2 and 4 movz/movk instructions (depending on the operand) Acked-by:
-
Ard Biesheuvel authored
Refactor the relocation processing so that the code executes from the ID map while accessing the relocation tables via the virtual mapping. This way, we can use literals containing virtual addresses as before, instead of having to use convoluted absolute expressions. For symmetry with the secondary code path, the relocation code and the subsequent jump to the virtual entry point are implemented in a function called __primary_switch(), and __mmap_switched() is renamed to __primary_switched(). Also, the call sequence in stext() is aligned with the one in secondary_startup(), by replacing the awkward 'adr_l lr' and 'b cpu_setup' sequence with a simple branch and link. Acked-by:
-
Ard Biesheuvel authored
We can simply use a relocated 64-bit literal to store the address of __secondary_switched(), and the relocation code will ensure that it holds the correct value at secondary entry time, as long as we make sure that the literal is not dereferenced until after we have enabled the MMU. So jump via a small __secondary_switch() function covered by the ID map that performs the literal load and branch-to-register. Acked-by:
-
Ard Biesheuvel authored
This unexports some symbols from head.S that are only used locally. Acked-by:
-
- 25 Apr, 2016 9 commits
-
-
Suzuki K Poulose authored
maxcpu=n sets the number of CPUs activated at boot time to a max of n, but allowing the remaining CPUs to be brought up later if the user decides to do so. However, on arm64 due to various reasons, we disallowed hotplugging CPUs beyond n, by marking them not present. Now that we have checks in place to make sure the hotplugged CPUs have compatible features with system and requires no new errata, relax the restriction. Cc: Will Deacon <will.deacon@arm.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: James Morse <james.morse@arm.com> Reviewed-by:
Suzuki K Poulose <suzuki.poulose@arm.com> Signed-off-by:
-
Suzuki K Poulose authored
CPU Errata work arounds are detected and applied to the kernel code at boot time and the data is then freed up. If a new hotplugged CPU requires a work around which was not applied at boot time, there is nothing we can do but simply fail the booting. Cc: Will Deacon <will.deacon@arm.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Andre Przywara <andre.przywara@arm.com> Reviewed-by:
-
Marc Zyngier authored
When introducing the whole CPU feature detection framework, we lost the capability to detect a mismatched GIC configuration (using the GICv2 MMIO interface, but having the system register interface enabled). In order to solve this, use the new this_cpu_has_cap() helper. Also move the check to the CPU interface path in order to catch systems where the first CPU has been correctly configured, but the secondaries are not. Reviewed-by:
-
Marc Zyngier authored
Now that the capabilities are only available once all the CPUs have booted, we're unable to check for a particular feature in any subsystem that gets initialized before then. In order to support this, introduce a local_cpu_has_cap() function that tests for the presence of a given capability independently of the whole framework. Reviewed-by:
-
Suzuki K Poulose authored
Add scope parameter to the arm64_cpu_capabilities::matches(), so that this can be reused for checking the capability on a given CPU vs the system wide. The system uses the default scope associated with the capability for initialising the CPU_HWCAPs and ELF_HWCAPs. Cc: James Morse <james.morse@arm.com> Cc: Marc Zyngier <marc.zyngier@arm.com> Cc: Andre Przywara <andre.przywara@arm.com> Cc: Will Deacon <will.deacon@arm.com> Reviewed-by:
-
Mark Rutland authored
Improve the readability of dt_scan_depth1_nodes by removing the nested conditionals. Signed-off-by:
Stefano Stabellini <sstabellini@kernel.org> Signed-off-by:
-
Shannon Zhao authored
When it's a Xen domain0 booting with ACPI, it will supply a /chosen and a /hypervisor node in DT. So check if it needs to enable ACPI. Signed-off-by:
Shannon Zhao <shannon.zhao@linaro.org> Reviewed-by:
Stefano Stabellini <stefano.stabellini@eu.citrix.com> Acked-by:
Hanjun Guo <hanjun.guo@linaro.org> Tested-by:
Julien Grall <julien.grall@arm.com> Acked-by:
-
Ard Biesheuvel authored
Annotate the KASAN shadow region with boundary markers, so that its mappings stand out in the page table dumper output. Acked-by:
-
Ard Biesheuvel authored
There is no need to initialize the vmemmap region boundaries dynamically, since they are compile time constants. So just add these constants to the global struct initializer, and drop the dynamic assignment and related code. Acked-by:
-
- 22 Apr, 2016 3 commits
-
-
Ard Biesheuvel authored
The open coded conversion from struct page address to virtual address in lowmem_page_address() involves an intermediate conversion step to pfn number/physical address. Since the placement of the struct page array relative to the linear mapping may be completely independent from the placement of physical RAM (as is that case for arm64 after commit dfd55ad8 'arm64: vmemmap: use virtual projection of linear region'), the conversion to physical address and back again should factor out of the equation, but unfortunately, the shifting and pointer arithmetic involved prevent this from happening, and the resulting calculation essentially subtracts the address of the start of physical memory and adds it back again, in a way that prevents the compiler from optimizing it away. Since the start of physical memory is not a build time constant on arm64, the resulting conversion involves an unnecessary memory access, which we would like to get rid of. So replace the open coded conversion with a call to page_to_virt(), and use the open coded conversion as its default definition, to be overriden by the architecture, if desired. The existing arch specific definitions of page_to_virt are all equivalent to this default definition, so by itself this patch is a no-op. Acked-by:
Andrew Morton <akpm@linux-foundation.org> Acked-by:
-
Ard Biesheuvel authored
To align with generic code and other architectures that expect the macro page_to_virt to produce an expression whose type is 'void*', drop the arch specific definition, which is never referenced anyway. Acked-by:
-
Ard Biesheuvel authored
To align with other architectures, the expression produced by expanding the macro page_to_virt() should be of type void*, since it returns a virtual address. Fix that, and also fix up an instance where page_to_virt was expected to return 'unsigned long', and drop another instance that was entirely unused (page_to_bus) Acked-by:
-
- 21 Apr, 2016 2 commits
-
-
Robin Murphy authored
With the IOMMU core now taking care of default domains for groups regardless of bus type, we can gleefully rip out this stop-gap, as slight recompense for having to expand the other one. Acked-by:
Robin Murphy <robin.murphy@arm.com> Signed-off-by:
-
Robin Murphy authored
PCI devices now suffer the same hiccup as platform devices, in that they get their DMA ops configured before they have been added to their bus, and thus before we know whether they have successfully registered with an IOMMU or not. Until the necessary driver core changes to reorder calls during device creation have been worked out, extend our delayed notifier trick onto the PCI bus so as to avoid broken DMA ops once IOMMUs get plugged into the PCI code. Acked-by:
-
