From c11eb222fd7d4db91196121dbf854178505d2751 Mon Sep 17 00:00:00 2001 From: Mischa Jonker Date: Thu, 26 Sep 2013 15:44:56 +0200 Subject: ARC: Handle zero-overhead-loop in unaligned access handler If a load or store is the last instruction in a zero-overhead-loop, and it's misaligned, the loop would execute only once. This fixes that problem. Signed-off-by: Mischa Jonker Signed-off-by: Vineet Gupta --- arch/arc/kernel/unaligned.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'arch/arc') diff --git a/arch/arc/kernel/unaligned.c b/arch/arc/kernel/unaligned.c index 28d170060747..7ff5b5c183bb 100644 --- a/arch/arc/kernel/unaligned.c +++ b/arch/arc/kernel/unaligned.c @@ -245,6 +245,12 @@ int misaligned_fixup(unsigned long address, struct pt_regs *regs, regs->status32 &= ~STATUS_DE_MASK; } else { regs->ret += state.instr_len; + + /* handle zero-overhead-loop */ + if ((regs->ret == regs->lp_end) && (regs->lp_count)) { + regs->ret = regs->lp_start; + regs->lp_count--; + } } return 0; -- cgit v1.2.3 From 0752adfda15f0eca9859a76da3db1800e129ad43 Mon Sep 17 00:00:00 2001 From: Vineet Gupta Date: Thu, 26 Sep 2013 18:50:40 +0530 Subject: ARC: Fix 32-bit wrap around in access_ok() Anton reported | LTP tests syscalls/process_vm_readv01 and process_vm_writev01 fail | similarly in one testcase test_iov_invalid -> lvec->iov_base. | Testcase expects errno EFAULT and return code -1, | but it gets return code 1 and ERRNO is 0 what means success. Essentially test case was passing a pointer of -1 which access_ok() was not catching. It was doing [@addr + @sz <= TASK_SIZE] which would pass for @addr == -1 Fixed that by rewriting as [@addr <= TASK_SIZE - @sz] Reported-by: Anton Kolesov Signed-off-by: Vineet Gupta --- arch/arc/include/asm/uaccess.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'arch/arc') diff --git a/arch/arc/include/asm/uaccess.h b/arch/arc/include/asm/uaccess.h index 32420824375b..30c9baffa96f 100644 --- a/arch/arc/include/asm/uaccess.h +++ b/arch/arc/include/asm/uaccess.h @@ -43,7 +43,7 @@ * Because it essentially checks if buffer end is within limit and @len is * non-ngeative, which implies that buffer start will be within limit too. * - * The reason for rewriting being, for majorit yof cases, @len is generally + * The reason for rewriting being, for majority of cases, @len is generally * compile time constant, causing first sub-expression to be compile time * subsumed. * @@ -53,7 +53,7 @@ * */ #define __user_ok(addr, sz) (((sz) <= TASK_SIZE) && \ - (((addr)+(sz)) <= get_fs())) + ((addr) <= (get_fs() - (sz)))) #define __access_ok(addr, sz) (unlikely(__kernel_ok) || \ likely(__user_ok((addr), (sz)))) -- cgit v1.2.3 From 6c00350b573c0bd3635436e43e8696951dd6e1b6 Mon Sep 17 00:00:00 2001 From: Vineet Gupta Date: Wed, 25 Sep 2013 16:53:32 +0530 Subject: ARC: Workaround spinlock livelock in SMP SystemC simulation Some ARC SMP systems lack native atomic R-M-W (LLOCK/SCOND) insns and can only use atomic EX insn (reg with mem) to build higher level R-M-W primitives. This includes a SystemC based SMP simulation model. So rwlocks need to use a protecting spinlock for atomic cmp-n-exchange operation to update reader(s)/writer count. The spinlock operation itself looks as follows: mov reg, 1 ; 1=locked, 0=unlocked retry: EX reg, [lock] ; load existing, store 1, atomically BREQ reg, 1, rety ; if already locked, retry In single-threaded simulation, SystemC alternates between the 2 cores with "N" insn each based scheduling. Additionally for insn with global side effect, such as EX writing to shared mem, a core switch is enforced too. Given that, 2 cores doing a repeated EX on same location, Linux often got into a livelock e.g. when both cores were fiddling with tasklist lock (gdbserver / hackbench) for read/write respectively as the sequence diagram below shows: core1 core2 -------- -------- 1. spin lock [EX r=0, w=1] - LOCKED 2. rwlock(Read) - LOCKED 3. spin unlock [ST 0] - UNLOCKED spin lock [EX r=0,w=1] - LOCKED -- resched core 1---- 5. spin lock [EX r=1] - ALREADY-LOCKED -- resched core 2---- 6. rwlock(Write) - READER-LOCKED 7. spin unlock [ST 0] 8. rwlock failed, retry again 9. spin lock [EX r=0, w=1] -- resched core 1---- 10 spinlock locked in #9, retry #5 11. spin lock [EX gets 1] -- resched core 2---- ... ... The fix was to unlock using the EX insn too (step 7), to trigger another SystemC scheduling pass which would let core1 proceed, eliding the livelock. Signed-off-by: Vineet Gupta --- arch/arc/include/asm/spinlock.h | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'arch/arc') diff --git a/arch/arc/include/asm/spinlock.h b/arch/arc/include/asm/spinlock.h index f158197ac5b0..b6a8c2dfbe6e 100644 --- a/arch/arc/include/asm/spinlock.h +++ b/arch/arc/include/asm/spinlock.h @@ -45,7 +45,14 @@ static inline int arch_spin_trylock(arch_spinlock_t *lock) static inline void arch_spin_unlock(arch_spinlock_t *lock) { - lock->slock = __ARCH_SPIN_LOCK_UNLOCKED__; + unsigned int tmp = __ARCH_SPIN_LOCK_UNLOCKED__; + + __asm__ __volatile__( + " ex %0, [%1] \n" + : "+r" (tmp) + : "r"(&(lock->slock)) + : "memory"); + smp_mb(); } -- cgit v1.2.3 From 55c2e26204276b27f2b7a63123b701c950e45d89 Mon Sep 17 00:00:00 2001 From: Uwe Kleine-König Date: Tue, 24 Sep 2013 23:05:37 +0200 Subject: ARC: Use clockevents_config_and_register over clockevents_register_device MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit clockevents_config_and_register is more clever and correct than doing it by hand; so use it. [vgupta: fixed build failure due to missing ; in patch] Signed-off-by: Uwe Kleine-König Signed-off-by: Vineet Gupta --- arch/arc/kernel/time.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) (limited to 'arch/arc') diff --git a/arch/arc/kernel/time.c b/arch/arc/kernel/time.c index 0e51e69cf30d..3fde7de3ea67 100644 --- a/arch/arc/kernel/time.c +++ b/arch/arc/kernel/time.c @@ -227,12 +227,9 @@ void __attribute__((weak)) arc_local_timer_setup(unsigned int cpu) { struct clock_event_device *clk = &per_cpu(arc_clockevent_device, cpu); - clockevents_calc_mult_shift(clk, arc_get_core_freq(), 5); - - clk->max_delta_ns = clockevent_delta2ns(ARC_TIMER_MAX, clk); clk->cpumask = cpumask_of(cpu); - - clockevents_register_device(clk); + clockevents_config_and_register(clk, arc_get_core_freq(), + 0, ARC_TIMER_MAX); /* * setup the per-cpu timer IRQ handler - for all cpus -- cgit v1.2.3 From 10469350e345599dfef3fa78a7c19fb230e674c1 Mon Sep 17 00:00:00 2001 From: Christian Ruppert Date: Wed, 2 Oct 2013 11:13:38 +0200 Subject: ARC: Fix signal frame management for SA_SIGINFO Previously, when a signal was registered with SA_SIGINFO, parameters 2 and 3 of the signal handler were written to registers r1 and r2 before the register set was saved. This led to corruption of these two registers after returning from the signal handler (the wrong values were restored). With this patch, registers are now saved before any parameters are passed, thus maintaining the processor state from before signal entry. Signed-off-by: Christian Ruppert Signed-off-by: Vineet Gupta --- arch/arc/kernel/signal.c | 25 +++++++++++++------------ 1 file changed, 13 insertions(+), 12 deletions(-) (limited to 'arch/arc') diff --git a/arch/arc/kernel/signal.c b/arch/arc/kernel/signal.c index ee6ef2f60a28..7e95e1a86510 100644 --- a/arch/arc/kernel/signal.c +++ b/arch/arc/kernel/signal.c @@ -101,7 +101,6 @@ SYSCALL_DEFINE0(rt_sigreturn) { struct rt_sigframe __user *sf; unsigned int magic; - int err; struct pt_regs *regs = current_pt_regs(); /* Always make any pending restarted system calls return -EINTR */ @@ -119,15 +118,16 @@ SYSCALL_DEFINE0(rt_sigreturn) if (!access_ok(VERIFY_READ, sf, sizeof(*sf))) goto badframe; - err = restore_usr_regs(regs, sf); - err |= __get_user(magic, &sf->sigret_magic); - if (err) + if (__get_user(magic, &sf->sigret_magic)) goto badframe; if (unlikely(is_do_ss_needed(magic))) if (restore_altstack(&sf->uc.uc_stack)) goto badframe; + if (restore_usr_regs(regs, sf)) + goto badframe; + /* Don't restart from sigreturn */ syscall_wont_restart(regs); @@ -190,6 +190,15 @@ setup_rt_frame(int signo, struct k_sigaction *ka, siginfo_t *info, if (!sf) return 1; + /* + * w/o SA_SIGINFO, struct ucontext is partially populated (only + * uc_mcontext/uc_sigmask) for kernel's normal user state preservation + * during signal handler execution. This works for SA_SIGINFO as well + * although the semantics are now overloaded (the same reg state can be + * inspected by userland: but are they allowed to fiddle with it ? + */ + err |= stash_usr_regs(sf, regs, set); + /* * SA_SIGINFO requires 3 args to signal handler: * #1: sig-no (common to any handler) @@ -213,14 +222,6 @@ setup_rt_frame(int signo, struct k_sigaction *ka, siginfo_t *info, magic = MAGIC_SIGALTSTK; } - /* - * w/o SA_SIGINFO, struct ucontext is partially populated (only - * uc_mcontext/uc_sigmask) for kernel's normal user state preservation - * during signal handler execution. This works for SA_SIGINFO as well - * although the semantics are now overloaded (the same reg state can be - * inspected by userland: but are they allowed to fiddle with it ? - */ - err |= stash_usr_regs(sf, regs, set); err |= __put_user(magic, &sf->sigret_magic); if (err) return err; -- cgit v1.2.3 From 5b24282846c064ee90d40fcb3a8f63b8e754fd28 Mon Sep 17 00:00:00 2001 From: Vineet Gupta Date: Thu, 10 Oct 2013 19:33:57 +0530 Subject: ARC: Ignore ptrace SETREGSET request for synthetic register "stop_pc" ARCompact TRAP_S insn used for breakpoints, commits before exception is taken (updating architectural PC). So ptregs->ret contains next-PC and not the breakpoint PC itself. This is different from other restartable exceptions such as TLB Miss where ptregs->ret has exact faulting PC. gdb needs to know exact-PC hence ARC ptrace GETREGSET provides for @stop_pc which returns ptregs->ret vs. EFA depending on the situation. However, writing stop_pc (SETREGSET request), which updates ptregs->ret doesn't makes sense stop_pc doesn't always correspond to that reg as described above. This was not an issue so far since user_regs->ret / user_regs->stop_pc had same value and both writing to ptregs->ret was OK, needless, but NOT broken, hence not observed. With gdb "jump", they diverge, and user_regs->ret updating ptregs is overwritten immediately with stop_pc, which this patch fixes. Reported-by: Anton Kolesov Signed-off-by: Vineet Gupta --- arch/arc/kernel/ptrace.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/arc') diff --git a/arch/arc/kernel/ptrace.c b/arch/arc/kernel/ptrace.c index 333238564b67..5d76706139dd 100644 --- a/arch/arc/kernel/ptrace.c +++ b/arch/arc/kernel/ptrace.c @@ -102,7 +102,7 @@ static int genregs_set(struct task_struct *target, REG_IGNORE_ONE(pad2); REG_IN_CHUNK(callee, efa, cregs); /* callee_regs[r25..r13] */ REG_IGNORE_ONE(efa); /* efa update invalid */ - REG_IN_ONE(stop_pc, &ptregs->ret); /* stop_pc: PC update */ + REG_IGNORE_ONE(stop_pc); /* PC updated via @ret */ return ret; } -- cgit v1.2.3