From c5cea06be060f38e5400d796e61cfc8c36e52924 Mon Sep 17 00:00:00 2001 From: Mark Rutland Date: Mon, 13 Jun 2016 11:15:14 +0100 Subject: [PATCH 1/7] arm64: fix dump_instr when PAN and UAO are in use If the kernel is set to show unhandled signals, and a user task does not handle a SIGILL as a result of an instruction abort, we will attempt to log the offending instruction with dump_instr before killing the task. We use dump_instr to log the encoding of the offending userspace instruction. However, dump_instr is also used to dump instructions from kernel space, and internally always switches to KERNEL_DS before dumping the instruction with get_user. When both PAN and UAO are in use, reading a user instruction via get_user while in KERNEL_DS will result in a permission fault, which leads to an Oops. As we have regs corresponding to the context of the original instruction abort, we can inspect this and only flip to KERNEL_DS if the original abort was taken from the kernel, avoiding this issue. At the same time, remove the redundant (and incorrect) comments regarding the order dump_mem and dump_instr are called in. Cc: Catalin Marinas Cc: James Morse Cc: Robin Murphy Cc: #4.6+ Signed-off-by: Mark Rutland Reported-by: Vladimir Murzin Tested-by: Vladimir Murzin Fixes: 57f4959bad0a154a ("arm64: kernel: Add support for User Access Override") Signed-off-by: Will Deacon --- arch/arm64/kernel/traps.c | 26 +++++++++++++------------- 1 file changed, 13 insertions(+), 13 deletions(-) diff --git a/arch/arm64/kernel/traps.c b/arch/arm64/kernel/traps.c index f7cf463107df..2a43012616b7 100644 --- a/arch/arm64/kernel/traps.c +++ b/arch/arm64/kernel/traps.c @@ -64,8 +64,7 @@ static void dump_mem(const char *lvl, const char *str, unsigned long bottom, /* * We need to switch to kernel mode so that we can use __get_user - * to safely read from kernel space. Note that we now dump the - * code first, just in case the backtrace kills us. + * to safely read from kernel space. */ fs = get_fs(); set_fs(KERNEL_DS); @@ -111,21 +110,12 @@ static void dump_backtrace_entry(unsigned long where) print_ip_sym(where); } -static void dump_instr(const char *lvl, struct pt_regs *regs) +static void __dump_instr(const char *lvl, struct pt_regs *regs) { unsigned long addr = instruction_pointer(regs); - mm_segment_t fs; char str[sizeof("00000000 ") * 5 + 2 + 1], *p = str; int i; - /* - * We need to switch to kernel mode so that we can use __get_user - * to safely read from kernel space. Note that we now dump the - * code first, just in case the backtrace kills us. - */ - fs = get_fs(); - set_fs(KERNEL_DS); - for (i = -4; i < 1; i++) { unsigned int val, bad; @@ -139,8 +129,18 @@ static void dump_instr(const char *lvl, struct pt_regs *regs) } } printk("%sCode: %s\n", lvl, str); +} - set_fs(fs); +static void dump_instr(const char *lvl, struct pt_regs *regs) +{ + if (!user_mode(regs)) { + mm_segment_t fs = get_fs(); + set_fs(KERNEL_DS); + __dump_instr(lvl, regs); + set_fs(fs); + } else { + __dump_instr(lvl, regs); + } } static void dump_backtrace(struct pt_regs *regs, struct task_struct *tsk) From bbb1681ee3653bdcfc6a4ba31902738118311fd4 Mon Sep 17 00:00:00 2001 From: Mark Rutland Date: Mon, 13 Jun 2016 17:57:02 +0100 Subject: [PATCH 2/7] arm64: mm: mark fault_info table const Unlike the debug_fault_info table, we never intentionally alter the fault_info table at runtime, and all derived pointers are treated as const currently. Make the table const so that it can be placed in .rodata and protected from unintentional writes, as we do for the syscall tables. Signed-off-by: Mark Rutland Cc: Catalin Marinas Cc: Will Deacon Signed-off-by: Will Deacon --- arch/arm64/mm/fault.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/arch/arm64/mm/fault.c b/arch/arm64/mm/fault.c index ba3fc12bd272..013e2cbe7924 100644 --- a/arch/arm64/mm/fault.c +++ b/arch/arm64/mm/fault.c @@ -441,7 +441,7 @@ static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs) return 1; } -static struct fault_info { +static const struct fault_info { int (*fn)(unsigned long addr, unsigned int esr, struct pt_regs *regs); int sig; int code; From f7a6c1492a2cb596952260a7d5bb0d61ca815173 Mon Sep 17 00:00:00 2001 From: Mark Salter Date: Tue, 7 Jun 2016 11:32:21 -0500 Subject: [PATCH 3/7] arm: pmu: Fix non-devicetree probing There is a problem in the non-devicetree PMU probing where some probe functions may get the number of supported events through smp_call_function_any() using the arm_pmu supported_cpus mask. But at the time the probe function is called, the supported_cpus mask is empty so the call fails. This patch makes sure the mask is set before calling the init function rather than after. Signed-off-by: Mark Salter Signed-off-by: Jeremy Linton Signed-off-by: Will Deacon --- drivers/perf/arm_pmu.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/perf/arm_pmu.c b/drivers/perf/arm_pmu.c index 1b8304e1efaa..140436a046c0 100644 --- a/drivers/perf/arm_pmu.c +++ b/drivers/perf/arm_pmu.c @@ -1010,8 +1010,8 @@ int arm_pmu_device_probe(struct platform_device *pdev, if (!ret) ret = init_fn(pmu); } else { - ret = probe_current_pmu(pmu, probe_table); cpumask_setall(&pmu->supported_cpus); + ret = probe_current_pmu(pmu, probe_table); } if (ret) { From 38b850a73034f075c4088e7511b36ebbef9dce00 Mon Sep 17 00:00:00 2001 From: Will Deacon Date: Thu, 2 Jun 2016 15:27:04 +0100 Subject: [PATCH 4/7] arm64: spinlock: order spin_{is_locked,unlock_wait} against local locks spin_is_locked has grown two very different use-cases: (1) [The sane case] API functions may require a certain lock to be held by the caller and can therefore use spin_is_locked as part of an assert statement in order to verify that the lock is indeed held. For example, usage of assert_spin_locked. (2) [The insane case] There are two locks, where a CPU takes one of the locks and then checks whether or not the other one is held before accessing some shared state. For example, the "optimized locking" in ipc/sem.c. In the latter case, the sequence looks like: spin_lock(&sem->lock); if (!spin_is_locked(&sma->sem_perm.lock)) /* Access shared state */ and requires that the spin_is_locked check is ordered after taking the sem->lock. Unfortunately, since our spinlocks are implemented using a LDAXR/STXR sequence, the read of &sma->sem_perm.lock can be speculated before the STXR and consequently return a stale value. Whilst this hasn't been seen to cause issues in practice, PowerPC fixed the same issue in 51d7d5205d33 ("powerpc: Add smp_mb() to arch_spin_is_locked()") and, although we did something similar for spin_unlock_wait in d86b8da04dfa ("arm64: spinlock: serialise spin_unlock_wait against concurrent lockers") that doesn't actually take care of ordering against local acquisition of a different lock. This patch adds an smp_mb() to the start of our arch_spin_is_locked and arch_spin_unlock_wait routines to ensure that the lock value is always loaded after any other locks have been taken by the current CPU. Reported-by: Peter Zijlstra Signed-off-by: Will Deacon --- arch/arm64/include/asm/spinlock.h | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/arch/arm64/include/asm/spinlock.h b/arch/arm64/include/asm/spinlock.h index fc9682bfe002..aac64d55cb22 100644 --- a/arch/arm64/include/asm/spinlock.h +++ b/arch/arm64/include/asm/spinlock.h @@ -31,6 +31,12 @@ static inline void arch_spin_unlock_wait(arch_spinlock_t *lock) unsigned int tmp; arch_spinlock_t lockval; + /* + * Ensure prior spin_lock operations to other locks have completed + * on this CPU before we test whether "lock" is locked. + */ + smp_mb(); + asm volatile( " sevl\n" "1: wfe\n" @@ -148,6 +154,7 @@ static inline int arch_spin_value_unlocked(arch_spinlock_t lock) static inline int arch_spin_is_locked(arch_spinlock_t *lock) { + smp_mb(); /* See arch_spin_unlock_wait */ return !arch_spin_value_unlocked(READ_ONCE(*lock)); } From 3a5facd09da848193f5bcb0dea098a298bc1a29d Mon Sep 17 00:00:00 2001 From: Will Deacon Date: Wed, 8 Jun 2016 15:10:57 +0100 Subject: [PATCH 5/7] arm64: spinlock: fix spin_unlock_wait for LSE atomics Commit d86b8da04dfa ("arm64: spinlock: serialise spin_unlock_wait against concurrent lockers") fixed spin_unlock_wait for LL/SC-based atomics under the premise that the LSE atomics (in particular, the LDADDA instruction) are indivisible. Unfortunately, these instructions are only indivisible when used with the -AL (full ordering) suffix and, consequently, the same issue can theoretically be observed with LSE atomics, where a later (in program order) load can be speculated before the write portion of the atomic operation. This patch fixes the issue by performing a CAS of the lock once we've established that it's unlocked, in much the same way as the LL/SC code. Fixes: d86b8da04dfa ("arm64: spinlock: serialise spin_unlock_wait against concurrent lockers") Signed-off-by: Will Deacon --- arch/arm64/include/asm/spinlock.h | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) diff --git a/arch/arm64/include/asm/spinlock.h b/arch/arm64/include/asm/spinlock.h index aac64d55cb22..d5c894253e73 100644 --- a/arch/arm64/include/asm/spinlock.h +++ b/arch/arm64/include/asm/spinlock.h @@ -43,13 +43,17 @@ static inline void arch_spin_unlock_wait(arch_spinlock_t *lock) "2: ldaxr %w0, %2\n" " eor %w1, %w0, %w0, ror #16\n" " cbnz %w1, 1b\n" + /* Serialise against any concurrent lockers */ ARM64_LSE_ATOMIC_INSN( /* LL/SC */ " stxr %w1, %w0, %2\n" -" cbnz %w1, 2b\n", /* Serialise against any concurrent lockers */ - /* LSE atomics */ " nop\n" -" nop\n") +" nop\n", + /* LSE atomics */ +" mov %w1, %w0\n" +" cas %w0, %w0, %2\n" +" eor %w1, %w1, %w0\n") +" cbnz %w1, 2b\n" : "=&r" (lockval), "=&r" (tmp), "+Q" (*lock) : : "memory"); From c56bdcac153e60d96a619a59c7981f2a78cba598 Mon Sep 17 00:00:00 2001 From: Will Deacon Date: Thu, 2 Jun 2016 18:40:07 +0100 Subject: [PATCH 6/7] arm64: spinlock: Ensure forward-progress in spin_unlock_wait Rather than wait until we observe the lock being free (which might never happen), we can also return from spin_unlock_wait if we observe that the lock is now held by somebody else, which implies that it was unlocked but we just missed seeing it in that state. Furthermore, in such a scenario there is no longer a need to write back the value that we loaded, since we know that there has been a lock hand-off, which is sufficient to publish any stores prior to the unlock_wait because the ARm architecture ensures that a Store-Release instruction is multi-copy atomic when observed by a Load-Acquire instruction. The litmus test is something like: AArch64 { 0:X1=x; 0:X3=y; 1:X1=y; 2:X1=y; 2:X3=x; } P0 | P1 | P2 ; MOV W0,#1 | MOV W0,#1 | LDAR W0,[X1] ; STR W0,[X1] | STLR W0,[X1] | LDR W2,[X3] ; DMB SY | | ; LDR W2,[X3] | | ; exists (0:X2=0 /\ 2:X0=1 /\ 2:X2=0) where P0 is doing spin_unlock_wait, P1 is doing spin_unlock and P2 is doing spin_lock. Signed-off-by: Will Deacon --- arch/arm64/include/asm/spinlock.h | 27 ++++++++++++++++++++++++--- 1 file changed, 24 insertions(+), 3 deletions(-) diff --git a/arch/arm64/include/asm/spinlock.h b/arch/arm64/include/asm/spinlock.h index d5c894253e73..e875a5a551d7 100644 --- a/arch/arm64/include/asm/spinlock.h +++ b/arch/arm64/include/asm/spinlock.h @@ -30,20 +30,39 @@ static inline void arch_spin_unlock_wait(arch_spinlock_t *lock) { unsigned int tmp; arch_spinlock_t lockval; + u32 owner; /* * Ensure prior spin_lock operations to other locks have completed * on this CPU before we test whether "lock" is locked. */ smp_mb(); + owner = READ_ONCE(lock->owner) << 16; asm volatile( " sevl\n" "1: wfe\n" "2: ldaxr %w0, %2\n" + /* Is the lock free? */ " eor %w1, %w0, %w0, ror #16\n" -" cbnz %w1, 1b\n" - /* Serialise against any concurrent lockers */ +" cbz %w1, 3f\n" + /* Lock taken -- has there been a subsequent unlock->lock transition? */ +" eor %w1, %w3, %w0, lsl #16\n" +" cbz %w1, 1b\n" + /* + * The owner has been updated, so there was an unlock->lock + * transition that we missed. That means we can rely on the + * store-release of the unlock operation paired with the + * load-acquire of the lock operation to publish any of our + * previous stores to the new lock owner and therefore don't + * need to bother with the writeback below. + */ +" b 4f\n" +"3:\n" + /* + * Serialise against any concurrent lockers by writing back the + * unlocked lock value + */ ARM64_LSE_ATOMIC_INSN( /* LL/SC */ " stxr %w1, %w0, %2\n" @@ -53,9 +72,11 @@ static inline void arch_spin_unlock_wait(arch_spinlock_t *lock) " mov %w1, %w0\n" " cas %w0, %w0, %2\n" " eor %w1, %w1, %w0\n") + /* Somebody else wrote to the lock, GOTO 10 and reload the value */ " cbnz %w1, 2b\n" +"4:" : "=&r" (lockval), "=&r" (tmp), "+Q" (*lock) - : + : "r" (owner) : "memory"); } From 0d15ef677839dab8313fbb86c007c3175b638d03 Mon Sep 17 00:00:00 2001 From: Daniel Thompson Date: Thu, 16 Jun 2016 16:51:52 +0100 Subject: [PATCH 7/7] arm64: kgdb: Match pstate size with gdbserver protocol Current versions of gdb do not interoperate cleanly with kgdb on arm64 systems because gdb and kgdb do not use the same register description. This patch modifies kgdb to work with recent releases of gdb (>= 7.8.1). Compatibility with gdb (after the patch is applied) is as follows: gdb-7.6 and earlier Ok gdb-7.7 series Works if user provides custom target description gdb-7.8(.0) Works if user provides custom target description gdb-7.8.1 and later Ok When commit 44679a4f142b ("arm64: KGDB: Add step debugging support") was introduced it was paired with a gdb patch that made an incompatible change to the gdbserver protocol. This patch was eventually merged into the gdb sources: https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=commit;h=a4d9ba85ec5597a6a556afe26b712e878374b9dd The change to the protocol was mostly made to simplify big-endian support inside the kernel gdb stub. Unfortunately the gdb project released gdb-7.7.x and gdb-7.8.0 before the protocol incompatibility was identified and reversed: https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=commit;h=bdc144174bcb11e808b4e73089b850cf9620a7ee This leaves us in a position where kgdb still uses the no-longer-used protocol; gdb-7.8.1, which restored the original behaviour, was released on 2014-10-29. I don't believe it is possible to detect/correct the protocol incompatiblity which means the kernel must take a view about which version of the gdb remote protocol is "correct". This patch takes the view that the original/current version of the protocol is correct and that version found in gdb-7.7.x and gdb-7.8.0 is anomalous. Signed-off-by: Daniel Thompson Signed-off-by: Will Deacon --- arch/arm64/include/asm/kgdb.h | 45 ++++++++++++++++++++++++++++------- arch/arm64/kernel/kgdb.c | 14 ++++++++++- 2 files changed, 50 insertions(+), 9 deletions(-) diff --git a/arch/arm64/include/asm/kgdb.h b/arch/arm64/include/asm/kgdb.h index f69f69c8120c..da84645525b9 100644 --- a/arch/arm64/include/asm/kgdb.h +++ b/arch/arm64/include/asm/kgdb.h @@ -38,25 +38,54 @@ extern int kgdb_fault_expected; #endif /* !__ASSEMBLY__ */ /* - * gdb is expecting the following registers layout. + * gdb remote procotol (well most versions of it) expects the following + * register layout. * * General purpose regs: * r0-r30: 64 bit * sp,pc : 64 bit - * pstate : 64 bit - * Total: 34 + * pstate : 32 bit + * Total: 33 + 1 * FPU regs: * f0-f31: 128 bit - * Total: 32 - * Extra regs * fpsr & fpcr: 32 bit - * Total: 2 + * Total: 32 + 2 * + * To expand a little on the "most versions of it"... when the gdb remote + * protocol for AArch64 was developed it depended on a statement in the + * Architecture Reference Manual that claimed "SPSR_ELx is a 32-bit register". + * and, as a result, allocated only 32-bits for the PSTATE in the remote + * protocol. In fact this statement is still present in ARM DDI 0487A.i. + * + * Unfortunately "is a 32-bit register" has a very special meaning for + * system registers. It means that "the upper bits, bits[63:32], are + * RES0.". RES0 is heavily used in the ARM architecture documents as a + * way to leave space for future architecture changes. So to translate a + * little for people who don't spend their spare time reading ARM architecture + * manuals, what "is a 32-bit register" actually means in this context is + * "is a 64-bit register but one with no meaning allocated to any of the + * upper 32-bits... *yet*". + * + * Perhaps then we should not be surprised that this has led to some + * confusion. Specifically a patch, influenced by the above translation, + * that extended PSTATE to 64-bit was accepted into gdb-7.7 but the patch + * was reverted in gdb-7.8.1 and all later releases, when this was + * discovered to be an undocumented protocol change. + * + * So... it is *not* wrong for us to only allocate 32-bits to PSTATE + * here even though the kernel itself allocates 64-bits for the same + * state. That is because this bit of code tells the kernel how the gdb + * remote protocol (well most versions of it) describes the register state. + * + * Note that if you are using one of the versions of gdb that supports + * the gdb-7.7 version of the protocol you cannot use kgdb directly + * without providing a custom register description (gdb can load new + * protocol descriptions at runtime). */ -#define _GP_REGS 34 +#define _GP_REGS 33 #define _FP_REGS 32 -#define _EXTRA_REGS 2 +#define _EXTRA_REGS 3 /* * general purpose registers size in bytes. * pstate is only 4 bytes. subtract 4 bytes diff --git a/arch/arm64/kernel/kgdb.c b/arch/arm64/kernel/kgdb.c index b67531a13136..b5f063e5eff7 100644 --- a/arch/arm64/kernel/kgdb.c +++ b/arch/arm64/kernel/kgdb.c @@ -58,7 +58,17 @@ struct dbg_reg_def_t dbg_reg_def[DBG_MAX_REG_NUM] = { { "x30", 8, offsetof(struct pt_regs, regs[30])}, { "sp", 8, offsetof(struct pt_regs, sp)}, { "pc", 8, offsetof(struct pt_regs, pc)}, - { "pstate", 8, offsetof(struct pt_regs, pstate)}, + /* + * struct pt_regs thinks PSTATE is 64-bits wide but gdb remote + * protocol disagrees. Therefore we must extract only the lower + * 32-bits. Look for the big comment in asm/kgdb.h for more + * detail. + */ + { "pstate", 4, offsetof(struct pt_regs, pstate) +#ifdef CONFIG_CPU_BIG_ENDIAN + + 4 +#endif + }, { "v0", 16, -1 }, { "v1", 16, -1 }, { "v2", 16, -1 }, @@ -128,6 +138,8 @@ sleeping_thread_to_gdb_regs(unsigned long *gdb_regs, struct task_struct *task) memset((char *)gdb_regs, 0, NUMREGBYTES); thread_regs = task_pt_regs(task); memcpy((void *)gdb_regs, (void *)thread_regs->regs, GP_REG_BYTES); + /* Special case for PSTATE (check comments in asm/kgdb.h for details) */ + dbg_get_reg(33, gdb_regs + GP_REG_BYTES, thread_regs); } void kgdb_arch_set_pc(struct pt_regs *regs, unsigned long pc)