- 06 Feb, 2018 14 commits
-
-
Robin Murphy authored
Similarly to x86, mitigate speculation past an access_ok() check by masking the pointer against the address limit before use. Even if we don't expect speculative writes per se, it is plausible that a CPU may still speculate at least as far as fetching a cache line for writing, hence we also harden put_user() and clear_user() for peace of mind. Signed-off-by:
Robin Murphy <robin.murphy@arm.com> Signed-off-by:
Will Deacon <will.deacon@arm.com> Signed-off-by:
Catalin Marinas <catalin.marinas@arm.com>
-
Robin Murphy authored
Currently, USER_DS represents an exclusive limit while KERNEL_DS is inclusive. In order to do some clever trickery for speculation-safe masking, we need them both to behave equivalently - there aren't enough bits to make KERNEL_DS exclusive, so we have precisely one option. This also happens to correct a longstanding false negative for a range ending on the very top byte of kernel memory. Mark Rutland points out that we've actually got the semantics of addresses vs. segments muddled up in most of the places we need to amend, so shuffle the {USER,KERNEL}_DS definitions around such that we can correct those properly instead of just pasting "-1"s everywhere. Signed-off-by: -
Robin Murphy authored
Provide an optimised, assembly implementation of array_index_mask_nospec() for arm64 so that the compiler is not in a position to transform the code in ways which affect its ability to inhibit speculation (e.g. by introducing conditional branches). This is similar to the sequence used by x86, modulo architectural differences in the carry/borrow flags. Reviewed-by:
Mark Rutland <mark.rutland@arm.com> Signed-off-by:
-
Will Deacon authored
For CPUs capable of data value prediction, CSDB waits for any outstanding predictions to architecturally resolve before allowing speculative execution to continue. Provide macros to expose it to the arch code. Reviewed-by:
-
Will Deacon authored
The identity map is mapped as both writeable and executable by the SWAPPER_MM_MMUFLAGS and this is relied upon by the kpti code to manage a synchronisation flag. Update the .pushsection flags to reflect the actual mapping attributes. Reported-by:
Marc Zyngier <marc.zyngier@arm.com> Signed-off-by:
-
Will Deacon authored
pte_to_phys lives in assembler.h and takes its destination register as the first argument. Move phys_to_pte out of head.S to sit with its counterpart and rejig it to follow the same calling convention. Signed-off-by:
-
Will Deacon authored
We don't fully understand the Cavium ThunderX erratum, but it appears that mapping the kernel as nG can lead to horrible consequences such as attempting to execute userspace from kernel context. Since kpti isn't enabled for these CPUs anyway, simplify the comment justifying the lack of post_ttbr_update_workaround in the exception trampoline. Signed-off-by:
-
Will Deacon authored
Since AArch64 assembly instructions take the destination register as their first operand, do the same thing for the phys_to_ttbr macro. Acked-by:
-
Marc Zyngier authored
Cavium ThunderX's erratum 27456 results in a corruption of icache entries that are loaded from memory that is mapped as non-global (i.e. ASID-tagged). As KPTI is based on memory being mapped non-global, let's prevent it from kicking in if this erratum is detected. Signed-off-by:
-
Will Deacon authored
Defaulting to global mappings for kernel space is generally good for performance and appears to be necessary for Cavium ThunderX. If we subsequently decide that we need to enable kpti, then we need to rewrite our existing page table entries to be non-global. This is fiddly, and made worse by the possible use of contiguous mappings, which require a strict break-before-make sequence. Since the enable callback runs on each online CPU from stop_machine context, we can have all CPUs enter the idmap, where secondaries can wait for the primary CPU to rewrite swapper with its MMU off. It's all fairly horrible, but at least it only runs once. Tested-by:
-
Will Deacon authored
Break-before-make is not needed when transitioning from Global to Non-Global mappings, provided that the contiguous hint is not being used. Signed-off-by:
-
Will Deacon authored
To allow systems which do not require kpti to continue running with global kernel mappings (which appears to be a requirement for Cavium ThunderX due to a CPU erratum), make the use of nG in the kernel page tables dependent on arm64_kernel_unmapped_at_el0(), which is resolved at runtime. Signed-off-by:
-
Shanker Donthineni authored
The ARM architecture defines the memory locations that are permitted to be accessed as the result of a speculative instruction fetch from an exception level for which all stages of translation are disabled. Specifically, the core is permitted to speculatively fetch from the 4KB region containing the current program counter 4K and next 4K. When translation is changed from enabled to disabled for the running exception level (SCTLR_ELn[M] changed from a value of 1 to 0), the Falkor core may errantly speculatively access memory locations outside of the 4KB region permitted by the architecture. The errant memory access may lead to one of the following unexpected behaviors. 1) A System Error Interrupt (SEI) being raised by the Falkor core due to the errant memory access attempting to access a region of memory that is protected by a slave-side memory protection unit. 2) Unpredictable device behavior due to a speculative read from device memory. This behavior may only occur if the instruction cache is disabled prior to or coincident with translation being changed from enabled to disabled. The conditions leading to this erratum will not occur when either of the following occur: 1) A higher exception level disables translation of a lower exception level (e.g. EL2 changing SCTLR_EL1[M] from a value of 1 to 0). 2) An exception level disabling its stage-1 translation if its stage-2 translation is enabled (e.g. EL1 changing SCTLR_EL1[M] from a value of 1 to 0 when HCR_EL2[VM] has a value of 1). To avoid the errant behavior, software must execute an ISB immediately prior to executing the MSR that will change SCTLR_ELn[M] from 1 to 0. Signed-off-by:Shanker Donthineni <shankerd@codeaurora.org> Signed-off-by:
-
Will Deacon authored
If the spinlock "next" ticket wraps around between the initial LDR and the cmpxchg in the LSE version of spin_trylock, then we can erroneously think that we have successfuly acquired the lock because we only check whether the next ticket return by the cmpxchg is equal to the owner ticket in our updated lock word. This patch fixes the issue by performing a full 32-bit check of the lock word when trying to determine whether or not the CASA instruction updated memory. Reported-by:
-
- 26 Jan, 2018 1 commit
-
-
Steve Capper authored
In cpu_do_switch_mm(.) with ARM64_SW_TTBR0_PAN=y we apply phys_to_ttbr to a value that already has an ASID inserted into the upper bits. For 52-bit PA configurations this then can give us TTBR0_EL1 registers that cause translation table walks to attempt to access non-zero PA[51:48] spuriously. Ultimately leading to a Synchronous External Abort on level 1 translation. This patch re-arranges the logic in cpu_do_switch_mm(.) such that phys_to_ttbr is called before the ASID is inserted into the TTBR0 value. Fixes: 6b88a32c ("arm64: kpti: Fix the interaction between ASID switching and software PAN") Acked-by:
Suzuki K Poulose <suzuki.poulose@arm.com> Tested-by:
Kristina Martsenko <kristina.martsenko@arm.com> Reviewed-by:
Steve Capper <steve.capper@arm.com> Signed-off-by:
-
- 23 Jan, 2018 4 commits
-
-
Jayachandran C authored
Whitelist Broadcom Vulcan/Cavium ThunderX2 processors in unmap_kernel_at_el0(). These CPUs are not vulnerable to CVE-2017-5754 and do not need KPTI when KASLR is off. Acked-by:
Jayachandran C <jnair@caviumnetworks.com> Signed-off-by:
-
Jayachandran C authored
Use PSCI based mitigation for speculative execution attacks targeting the branch predictor. We use the same mechanism as the one used for Cortex-A CPUs, we expect the PSCI version call to have a side effect of clearing the BTBs. Acked-by:
-
Suzuki K Poulose authored
When a CPU is brought up after we have finalised the system wide capabilities (i.e, features and errata), we make sure the new CPU doesn't need a new errata work around which has not been detected already. However we don't run enable() method on the new CPU for the errata work arounds already detected. This could cause the new CPU running without potential work arounds. It is upto the "enable()" method to decide if this CPU should do something about the errata. Fixes: commit 6a6efbb4 ("arm64: Verify CPU errata work arounds on hotplugged CPU") Cc: Will Deacon <will.deacon@arm.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Andre Przywara <andre.przywara@arm.com> Cc: Dave Martin <dave.martin@arm.com> Signed-off-by:
-
Marc Zyngier authored
We call arm64_apply_bp_hardening() from post_ttbr_update_workaround, which has the unexpected consequence of being triggered on every exception return to userspace when ARM64_SW_TTBR0_PAN is selected, even if no context switch actually occured. This is a bit suboptimal, and it would be more logical to only invalidate the branch predictor when we actually switch to a different mm. In order to solve this, move the call to arm64_apply_bp_hardening() into check_and_switch_context(), where we're guaranteed to pick a different mm context. Acked-by:
-
- 19 Jan, 2018 1 commit
-
-
Kristina Martsenko authored
When booting a kernel without 52-bit PA support (e.g. a kernel with 4k pages) on a system with 52-bit memory, the kernel will currently try to use the 52-bit memory and crash. Fix this by ignoring any memory higher than what the kernel supports. Fixes: f77d2817 ("arm64: enable 52-bit physical address support") Signed-off-by:
-
- 16 Jan, 2018 17 commits
-
-
Catalin Marinas authored
With ARM64_SW_TTBR0_PAN enabled, the exception entry code checks the active ASID to decide whether user access was enabled (non-zero ASID) when the exception was taken. On return from exception, if user access was previously disabled, it re-instates TTBR0_EL1 from the per-thread saved value (updated in switch_mm() or efi_set_pgd()). Commit 7655abb9 ("arm64: mm: Move ASID from TTBR0 to TTBR1") makes a TTBR0_EL1 + ASID switching non-atomic. Subsequently, commit 27a921e7 ("arm64: mm: Fix and re-enable ARM64_SW_TTBR0_PAN") changes the __uaccess_ttbr0_disable() function and asm macro to first write the reserved TTBR0_EL1 followed by the ASID=0 update in TTBR1_EL1. If an exception occurs between these two, the exception return code will re-instate a valid TTBR0_EL1. Similar scenario can happen in cpu_switch_mm() between setting the reserved TTBR0_EL1 and the ASID update in cpu_do_switch_mm(). This patch reverts the entry.S check for ASID == 0 to TTBR0_EL1 and disables the interrupts around the TTBR0_EL1 and ASID switching code in __uaccess_ttbr0_disable(). It also ensures that, when returning from the EFI runtime services, efi_set_pgd() doesn't leave a non-zero ASID in TTBR1_EL1 by using uaccess_ttbr0_{enable,disable}. The accesses to current_thread_info()->ttbr0 are updated to use READ_ONCE/WRITE_ONCE. As a safety measure, __uaccess_ttbr0_enable() always masks out any existing non-zero ASID TTBR1_EL1 before writing in the new ASID. Fixes: 27a921e7 ("arm64: mm: Fix and re-enable ARM64_SW_TTBR0_PAN") Acked-by:
Ard Biesheuvel <ard.biesheuvel@linaro.org> Tested-by:
James Morse <james.morse@arm.com> Tested-by:
-
Dongjiu Geng authored
ARMv8.2 adds a new bit HCR_EL2.TEA which routes synchronous external aborts to EL2, and adds a trap control bit HCR_EL2.TERR which traps all Non-secure EL1&0 error record accesses to EL2. This patch enables the two bits for the guest OS, guaranteeing that KVM takes external aborts and traps attempts to access the physical error registers. ERRIDR_EL1 advertises the number of error records, we return zero meaning we can treat all the other registers as RAZ/WI too. Signed-off-by:
Dongjiu Geng <gengdongjiu@huawei.com> [removed specific emulation, use trap_raz_wi() directly for everything, rephrased parts of the commit message] Signed-off-by:
Christoffer Dall <christoffer.dall@linaro.org> Reviewed-by:
-
James Morse authored
We expect to have firmware-first handling of RAS SErrors, with errors notified via an APEI method. For systems without firmware-first, add some minimal handling to KVM. There are two ways KVM can take an SError due to a guest, either may be a RAS error: we exit the guest due to an SError routed to EL2 by HCR_EL2.AMO, or we take an SError from EL2 when we unmask PSTATE.A from __guest_exit. The current SError from EL2 code unmasks SError and tries to fence any pending SError into a single instruction window. It then leaves SError unmasked. With the v8.2 RAS Extensions we may take an SError for a 'corrected' error, but KVM is only able to handle SError from EL2 if they occur during this single instruction window... The RAS Extensions give us a new instruction to synchronise and consume SErrors. The RAS Extensions document (ARM DDI0587), '2.4.1 ESB and Unrecoverable errors' describes ESB as synchronising SError interrupts generated by 'instructions, translation table walks, hardware updates to the translation tables, and instruction fetches on the same PE'. This makes ESB equivalent to KVMs existing 'dsb, mrs-daifclr, isb' sequence. Use the alternatives to synchronise and consume any SError using ESB instead of unmasking and taking the SError. Set ARM_EXIT_WITH_SERROR_BIT in the exit_code so that we can restart the vcpu if it turns out this SError has no impact on the vcpu. Reviewed-by:
-
James Morse authored
We expect to have firmware-first handling of RAS SErrors, with errors notified via an APEI method. For systems without firmware-first, add some minimal handling to KVM. There are two ways KVM can take an SError due to a guest, either may be a RAS error: we exit the guest due to an SError routed to EL2 by HCR_EL2.AMO, or we take an SError from EL2 when we unmask PSTATE.A from __guest_exit. For SError that interrupt a guest and are routed to EL2 the existing behaviour is to inject an impdef SError into the guest. Add code to handle RAS SError based on the ESR. For uncontained and uncategorized errors arm64_is_fatal_ras_serror() will panic(), these errors compromise the host too. All other error types are contained: For the fatal errors the vCPU can't make progress, so we inject a virtual SError. We ignore contained errors where we can make progress as if we're lucky, we may not hit them again. If only some of the CPUs support RAS the guest will see the cpufeature sanitised version of the id registers, but we may still take RAS SError on this CPU. Move the SError handling out of handle_exit() into a new handler that runs before we can be preempted. This allows us to use this_cpu_has_cap(), via arm64_is_ras_serror(). Acked-by:
-
James Morse authored
When we exit a guest due to an SError the vcpu fault info isn't updated with the ESR. Today this is only done for traps. The v8.2 RAS Extensions define ISS values for SError. Update the vcpu's fault_info with the ESR on SError so that handle_exit() can determine if this was a RAS SError and decode its severity. Acked-by:
-
James Morse authored
If we deliver a virtual SError to the guest, the guest may defer it with an ESB instruction. The guest reads the deferred value via DISR_EL1, but the guests view of DISR_EL1 is re-mapped to VDISR_EL2 when HCR_EL2.AMO is set. Add the KVM code to save/restore VDISR_EL2, and make it accessible to userspace as DISR_EL1. Signed-off-by:
-
James Morse authored
Prior to v8.2's RAS Extensions, the HCR_EL2.VSE 'virtual SError' feature generated an SError with an implementation defined ESR_EL1.ISS, because we had no mechanism to specify the ESR value. On Juno this generates an all-zero ESR, the most significant bit 'ISV' is clear indicating the remainder of the ISS field is invalid. With the RAS Extensions we have a mechanism to specify this value, and the most significant bit has a new meaning: 'IDS - Implementation Defined Syndrome'. An all-zero SError ESR now means: 'RAS error: Uncategorized' instead of 'no valid ISS'. Add KVM support for the VSESR_EL2 register to specify an ESR value when HCR_EL2.VSE generates a virtual SError. Change kvm_inject_vabt() to specify an implementation-defined value. We only need to restore the VSESR_EL2 value when HCR_EL2.VSE is set, KVM save/restores this bit during __{,de}activate_traps() and hardware clears the bit once the guest has consumed the virtual-SError. Future patches may add an API (or KVM CAP) to pend a virtual SError with a specified ESR. Cc: Dongjiu Geng <gengdongjiu@huawei.com> Reviewed-by: -
James Morse authored
Non-VHE systems take an exception to EL2 in order to world-switch into the guest. When returning from the guest KVM implicitly restores the DAIF flags when it returns to the kernel at EL1. With VHE none of this exception-level jumping happens, so KVMs world-switch code is exposed to the host kernel's DAIF values, and KVM spills the guest-exit DAIF values back into the host kernel. On entry to a guest we have Debug and SError exceptions unmasked, KVM has switched VBAR but isn't prepared to handle these. On guest exit Debug exceptions are left disabled once we return to the host and will stay this way until we enter user space. Add a helper to mask/unmask DAIF around VHE guests. The unmask can only happen after the hosts VBAR value has been synchronised by the isb in __vhe_hyp_call (via kvm_call_hyp()). Masking could be as late as setting KVMs VBAR value, but is kept here for symmetry. Acked-by:
-
James Morse authored
KVM would like to consume any pending SError (or RAS error) after guest exit. Today it has to unmask SError and use dsb+isb to synchronise the CPU. With the RAS extensions we can use ESB to synchronise any pending SError. Add the necessary macros to allow DISR to be read and converted to an ESR. We clear the DISR register when we enable the RAS cpufeature, and the kernel has not executed any ESB instructions. Any value we find in DISR must have belonged to firmware. Executing an ESB instruction is the only way to update DISR, so we can expect firmware to have handled any deferred SError. By the same logic we clear DISR in the idle path. Reviewed-by:
-
James Morse authored
ARM v8.2 has a feature to add implicit error synchronization barriers whenever the CPU enters or returns from an exception level. Add this to the features we always enable. CPUs that don't support this feature will treat the bit as RES0. This feature causes RAS errors that are not yet visible to software to become pending SErrors. We expect to have firmware-first RAS support so synchronised RAS errors will be take immediately to EL3. Any system without firmware-first handling of errors will take the SError either immediatly after exception return, or when we unmask SError after entry.S's work. Adding IESB to the ELx flags causes it to be enabled by KVM and kexec too. Platform level RAS support may require additional firmware support. Cc: Christoffer Dall <christoffer.dall@linaro.org> Suggested-by:
-
James Morse authored
Prior to v8.2, SError is an uncontainable fatal exception. The v8.2 RAS extensions use SError to notify software about RAS errors, these can be contained by the Error Syncronization Barrier. An ACPI system with firmware-first may use SError as its 'SEI' notification. Future patches may add code to 'claim' this SError as a notification. Other systems can distinguish these RAS errors from the SError ESR and use the AET bits and additional data from RAS-Error registers to handle the error. Future patches may add this kernel-first handling. Without support for either of these we will panic(), even if we received a corrected error. Add code to decode the severity of RAS errors. We can safely ignore contained errors where the CPU can continue to make progress. For all other errors we continue to panic(). Signed-off-by:
-
Xie XiuQi authored
ARM's v8.2 Extentions add support for Reliability, Availability and Serviceability (RAS). On CPUs with these extensions system software can use additional barriers to isolate errors and determine if faults are pending. Add cpufeature detection. Platform level RAS support may require additional firmware support. Reviewed-by:
Xie XiuQi <xiexiuqi@huawei.com> [Rebased added config option, reworded commit message] Signed-off-by:
-
James Morse authored
__cpu_setup() configures SCTLR_EL1 using some hard coded hex masks, and el2_setup() duplicates some this when setting RES1 bits. Lets make this the same as KVM's hyp_init, which uses named bits. First, we add definitions for all the SCTLR_EL{1,2} bits, the RES{1,0} bits, and those we want to set or clear. Add a build_bug checks to ensures all bits are either set or clear. This means we don't need to preserve endian-ness configuration generated elsewhere. Finally, move the head.S and proc.S users of these hard-coded masks over to the macro versions. Signed-off-by: -
James Morse authored
this_cpu_has_cap() tests caps->desc not caps->matches, so it stops walking the list when it finds a 'silent' feature, instead of walking to the end of the list. Prior to v4.6's 644c2ae1 ("arm64: cpufeature: Test 'matches' pointer to find the end of the list") we always tested desc to find the end of a capability list. This was changed for dubious things like PAN_NOT_UAO. v4.7's e3661b12 ("arm64: Allow a capability to be checked on single CPU") added this_cpu_has_cap() using the old desc style test. CC: Suzuki K Poulose <suzuki.poulose@arm.com> Reviewed-by:
-
Dave Martin authored
When refactoring the sigreturn code to handle SVE, I changed the sigreturn implementation to store the new FPSIMD state from the user sigframe into task_struct before reloading the state into the CPU regs. This makes it easier to convert the data for SVE when needed. However, it turns out that the fpsimd_state structure passed into fpsimd_update_current_state is not fully initialised, so assigning the structure as a whole corrupts current->thread.fpsimd_state.cpu with uninitialised data. This means that if the garbage data written to .cpu happens to be a valid cpu number, and the task is subsequently migrated to the cpu identified by the that number, and then tries to enter userspace, the CPU FPSIMD regs will be assumed to be correct for the task and not reloaded as they should be. This can result in returning to userspace with the FPSIMD registers containing data that is stale or that belongs to another task or to the kernel. Knowingly handing around a kernel structure that is incompletely initialised with user data is a potential source of mistakes, especially across source file boundaries. To help avoid a repeat of this issue, this patch adapts the relevant internal API to hand around the user-accessible subset only: struct user_fpsimd_state. To avoid future surprises, this patch also converts all uses of struct fpsimd_state that really only access the user subset, to use struct user_fpsimd_state. A few missing consts are added to function prototypes for good measure. Thanks to Will for spotting the cause of the bug here. Reported-by:
Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by:
Dave Martin <Dave.Martin@arm.com> Cc: Will Deacon <will.deacon@arm.com> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by:
-
Punit Agrawal authored
The PUD macros (PUD_TABLE_BIT, PUD_TYPE_MASK, PUD_TYPE_SECT) use the pgdval_t even when pudval_t is available. Even though the underlying type for both (u64) is the same it is confusing and may lead to issues in the future. Fix this by using pudval_t to define the PUD_* macros. Fixes: 084bd298 ("ARM64: mm: HugeTLB support.") Fixes: 206a2a73 ("arm64: mm: Create gigabyte kernel logical mappings where possible") Signed-off-by:
Punit Agrawal <punit.agrawal@arm.com> Cc: Will Deacon <will.deacon@arm.com> Signed-off-by:
-
Stephen Boyd authored
It isn't entirely obvious if we're using software PAN because we don't say anything about it in the boot log. But if we're using hardware PAN we'll print a nice CPU feature message indicating it. Add a print for software PAN too so we know if it's being used or not. Signed-off-by:
Stephen Boyd <sboyd@codeaurora.org> Signed-off-by:
-
- 15 Jan, 2018 3 commits
-
-
Kristina Martsenko authored
The 'pos' argument is used to select where in TCR to write the value: the IPS or PS bitfield. Fixes: 787fd1d0 ("arm64: limit PA size to supported range") Signed-off-by:
-
Kristina Martsenko authored
Commit fa2a8445 incorrectly masks the index of the HYP ID map pgd entry, causing a non-VHE kernel to hang during boot. This happens when VA_BITS=48 and the ID map text is in 52-bit physical memory. In this case we don't need an extra table level but need more entries in the top-level table, so we need to map into hyp_pgd and need to use __kvm_idmap_ptrs_per_pgd to mask in the extra bits. However, __create_hyp_mappings currently masks by PTRS_PER_PGD instead. Fix it so that we always use __kvm_idmap_ptrs_per_pgd for the HYP ID map. This ensures that we use the larger mask for the top-level ID map table when it has more entries. In all other cases, PTRS_PER_PGD is used as normal. Fixes: fa2a8445 ("arm64: allow ID map to be extended to 52 bits") Acked-by:
-
Kristina Martsenko authored
Commit fa2a8445 added support for extending the ID map to 52 bits, but accidentally dropped a required change to __cpu_uses_extended_idmap. As a result, the kernel fails to boot when VA_BITS = 48 and the ID map text is in 52-bit physical memory, because we reduce TCR.T0SZ to cover the ID map, but then never set it back to VA_BITS. Add back the change, and also clean up some double parentheses. Fixes: fa2a8445 ("arm64: allow ID map to be extended to 52 bits") Reviewed-by:
-
