diff options
author | Andi Kleen <andi@firstfloor.org> | 2009-02-12 13:43:23 +0100 |
---|---|---|
committer | H. Peter Anvin <hpa@linux.intel.com> | 2009-02-19 14:52:20 -0800 |
commit | b79109c3bbcf52cac5103979b283b9e5df4e796c (patch) | |
tree | 528da2232a2c7f013bde63b623bc11a205ce8773 | |
parent | b5f2fa4ea00a179ac1c2ff342ceeee261dd75e53 (diff) | |
download | linux-b79109c3bbcf52cac5103979b283b9e5df4e796c.tar.bz2 |
x86, mce: separate correct machine check poller and fatal exception handler
Impact: cleanup, performance enhancement
The machine check poller is diverging more and more from the fatal
exception handler. Instead of adding more special cases separate the code
paths completely. The corrected poll path is actually quite simple,
and this doesn't result in much code duplication.
This makes both handlers much easier to read and results in
cleaner code flow. The exception handler now only needs to care
about uncorrected errors, which also simplifies the handling of multiple
errors. The corrected poller also now always runs in standard interrupt
context and does not need to do anything special to handle NMI context.
Minor behaviour changes:
- MCG status is now not cleared on polling.
- Only the banks which had corrected errors get cleared on polling
- The exception handler only clears banks with errors now
v2: Forward port to new patch order. Add "uc" argument.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
-rw-r--r-- | arch/x86/include/asm/mce.h | 7 | ||||
-rw-r--r-- | arch/x86/kernel/cpu/mcheck/mce_64.c | 129 | ||||
-rw-r--r-- | arch/x86/kernel/cpu/mcheck/mce_amd_64.c | 2 |
3 files changed, 116 insertions, 22 deletions
diff --git a/arch/x86/include/asm/mce.h b/arch/x86/include/asm/mce.h index 048b71d9387a..225cdb5d2bfc 100644 --- a/arch/x86/include/asm/mce.h +++ b/arch/x86/include/asm/mce.h @@ -112,6 +112,13 @@ void mce_log_therm_throt_event(__u64 status); extern atomic_t mce_entry; extern void do_machine_check(struct pt_regs *, long); + +enum mcp_flags { + MCP_TIMESTAMP = (1 << 0), /* log time stamp */ + MCP_UC = (1 << 1), /* log uncorrected errors */ +}; +extern void machine_check_poll(enum mcp_flags flags); + extern int mce_notify_user(void); #endif /* !CONFIG_X86_32 */ diff --git a/arch/x86/kernel/cpu/mcheck/mce_64.c b/arch/x86/kernel/cpu/mcheck/mce_64.c index fed875742b1a..268b05edade7 100644 --- a/arch/x86/kernel/cpu/mcheck/mce_64.c +++ b/arch/x86/kernel/cpu/mcheck/mce_64.c @@ -3,6 +3,8 @@ * K8 parts Copyright 2002,2003 Andi Kleen, SuSE Labs. * Rest from unknown author(s). * 2004 Andi Kleen. Rewrote most of it. + * Copyright 2008 Intel Corporation + * Author: Andi Kleen */ #include <linux/init.h> @@ -189,7 +191,77 @@ static inline void mce_get_rip(struct mce *m, struct pt_regs *regs) } /* - * The actual machine check handler + * Poll for corrected events or events that happened before reset. + * Those are just logged through /dev/mcelog. + * + * This is executed in standard interrupt context. + */ +void machine_check_poll(enum mcp_flags flags) +{ + struct mce m; + int i; + + mce_setup(&m); + + rdmsrl(MSR_IA32_MCG_STATUS, m.mcgstatus); + for (i = 0; i < banks; i++) { + if (!bank[i]) + continue; + + m.misc = 0; + m.addr = 0; + m.bank = i; + m.tsc = 0; + + barrier(); + rdmsrl(MSR_IA32_MC0_STATUS + i*4, m.status); + if (!(m.status & MCI_STATUS_VAL)) + continue; + + /* + * Uncorrected events are handled by the exception handler + * when it is enabled. But when the exception is disabled log + * everything. + * + * TBD do the same check for MCI_STATUS_EN here? + */ + if ((m.status & MCI_STATUS_UC) && !(flags & MCP_UC)) + continue; + + if (m.status & MCI_STATUS_MISCV) + rdmsrl(MSR_IA32_MC0_MISC + i*4, m.misc); + if (m.status & MCI_STATUS_ADDRV) + rdmsrl(MSR_IA32_MC0_ADDR + i*4, m.addr); + + if (!(flags & MCP_TIMESTAMP)) + m.tsc = 0; + /* + * Don't get the IP here because it's unlikely to + * have anything to do with the actual error location. + */ + + mce_log(&m); + add_taint(TAINT_MACHINE_CHECK); + + /* + * Clear state for this bank. + */ + wrmsrl(MSR_IA32_MC0_STATUS+4*i, 0); + } + + /* + * Don't clear MCG_STATUS here because it's only defined for + * exceptions. + */ +} + +/* + * The actual machine check handler. This only handles real + * exceptions when something got corrupted coming in through int 18. + * + * This is executed in NMI context not subject to normal locking rules. This + * implies that most kernel services cannot be safely used. Don't even + * think about putting a printk in there! */ void do_machine_check(struct pt_regs * regs, long error_code) { @@ -207,13 +279,14 @@ void do_machine_check(struct pt_regs * regs, long error_code) * error. */ int kill_it = 0; + DECLARE_BITMAP(toclear, MAX_NR_BANKS); atomic_inc(&mce_entry); - if ((regs - && notify_die(DIE_NMI, "machine check", regs, error_code, + if (notify_die(DIE_NMI, "machine check", regs, error_code, 18, SIGKILL) == NOTIFY_STOP) - || !banks) + goto out2; + if (!banks) goto out2; mce_setup(&m); @@ -227,6 +300,7 @@ void do_machine_check(struct pt_regs * regs, long error_code) barrier(); for (i = 0; i < banks; i++) { + __clear_bit(i, toclear); if (!bank[i]) continue; @@ -238,6 +312,20 @@ void do_machine_check(struct pt_regs * regs, long error_code) if ((m.status & MCI_STATUS_VAL) == 0) continue; + /* + * Non uncorrected errors are handled by machine_check_poll + * Leave them alone. + */ + if ((m.status & MCI_STATUS_UC) == 0) + continue; + + /* + * Set taint even when machine check was not enabled. + */ + add_taint(TAINT_MACHINE_CHECK); + + __set_bit(i, toclear); + if (m.status & MCI_STATUS_EN) { /* if PCC was set, there's no way out */ no_way_out |= !!(m.status & MCI_STATUS_PCC); @@ -251,6 +339,12 @@ void do_machine_check(struct pt_regs * regs, long error_code) no_way_out = 1; kill_it = 1; } + } else { + /* + * Machine check event was not enabled. Clear, but + * ignore. + */ + continue; } if (m.status & MCI_STATUS_MISCV) @@ -259,10 +353,7 @@ void do_machine_check(struct pt_regs * regs, long error_code) rdmsrl(MSR_IA32_MC0_ADDR + i*4, m.addr); mce_get_rip(&m, regs); - if (error_code < 0) - m.tsc = 0; - if (error_code != -2) - mce_log(&m); + mce_log(&m); /* Did this bank cause the exception? */ /* Assume that the bank with uncorrectable errors did it, @@ -271,14 +362,8 @@ void do_machine_check(struct pt_regs * regs, long error_code) panicm = m; panicm_found = 1; } - - add_taint(TAINT_MACHINE_CHECK); } - /* Never do anything final in the polling timer */ - if (!regs) - goto out; - /* If we didn't find an uncorrectable error, pick the last one (shouldn't happen, just being safe). */ if (!panicm_found) @@ -325,10 +410,11 @@ void do_machine_check(struct pt_regs * regs, long error_code) /* notify userspace ASAP */ set_thread_flag(TIF_MCE_NOTIFY); - out: /* the last thing we do is clear state */ - for (i = 0; i < banks; i++) - wrmsrl(MSR_IA32_MC0_STATUS+4*i, 0); + for (i = 0; i < banks; i++) { + if (test_bit(i, toclear)) + wrmsrl(MSR_IA32_MC0_STATUS+4*i, 0); + } wrmsrl(MSR_IA32_MCG_STATUS, 0); out2: atomic_dec(&mce_entry); @@ -377,7 +463,7 @@ static void mcheck_timer(unsigned long data) WARN_ON(smp_processor_id() != data); if (mce_available(¤t_cpu_data)) - do_machine_check(NULL, 0); + machine_check_poll(MCP_TIMESTAMP); /* * Alert userspace if needed. If we logged an MCE, reduce the @@ -494,9 +580,10 @@ static void mce_init(void *dummy) u64 cap; int i; - /* Log the machine checks left over from the previous reset. - This also clears all registers */ - do_machine_check(NULL, mce_bootlog ? -1 : -2); + /* + * Log the machine checks left over from the previous reset. + */ + machine_check_poll(MCP_UC); set_in_cr4(X86_CR4_MCE); diff --git a/arch/x86/kernel/cpu/mcheck/mce_amd_64.c b/arch/x86/kernel/cpu/mcheck/mce_amd_64.c index 75d9dd25e3dc..0069c653f4ed 100644 --- a/arch/x86/kernel/cpu/mcheck/mce_amd_64.c +++ b/arch/x86/kernel/cpu/mcheck/mce_amd_64.c @@ -231,7 +231,7 @@ asmlinkage void mce_threshold_interrupt(void) /* Log the machine check that caused the threshold event. */ - do_machine_check(NULL, 0); + machine_check_poll(MCP_TIMESTAMP); if (high & MASK_OVERFLOW_HI) { rdmsrl(address, m.misc); |