From: Borislav Petkov bp@suse.de
commit 487d654db3edacc31dee86b10258cc740640fad8 upstream.
Instead, sandwitch around the call which is done in noinstr context and mark the caller - mce_gather_info() - as noinstr.
Also, document what the whole instrumentation strategy with #MC is going to be in the future and where it all is supposed to be going to.
Signed-off-by: Borislav Petkov bp@suse.de Link: https://lore.kernel.org/r/20211208111343.8130-5-bp@alien8.de Signed-off-by: Greg Kroah-Hartman gregkh@linuxfoundation.org --- arch/x86/kernel/cpu/mce/core.c | 26 ++++++++++++++++++++------ 1 file changed, 20 insertions(+), 6 deletions(-)
--- a/arch/x86/kernel/cpu/mce/core.c +++ b/arch/x86/kernel/cpu/mce/core.c @@ -130,7 +130,7 @@ static void (*quirk_no_way_out)(int bank BLOCKING_NOTIFIER_HEAD(x86_mce_decoder_chain);
/* Do initial initialization of a struct mce */ -noinstr void mce_setup(struct mce *m) +void mce_setup(struct mce *m) { memset(m, 0, sizeof(struct mce)); m->cpu = m->extcpu = smp_processor_id(); @@ -479,9 +479,15 @@ static noinstr void mce_wrmsrl(u32 msr, * check into our "mce" struct so that we can use it later to assess * the severity of the problem as we read per-bank specific details. */ -static inline void mce_gather_info(struct mce *m, struct pt_regs *regs) +static noinstr void mce_gather_info(struct mce *m, struct pt_regs *regs) { + /* + * Enable instrumentation around mce_setup() which calls external + * facilities. + */ + instrumentation_begin(); mce_setup(m); + instrumentation_end();
m->mcgstatus = mce_rdmsrl(MSR_IA32_MCG_STATUS); if (regs) { @@ -1327,11 +1333,11 @@ static void queue_task_work(struct mce * }
/* - * The actual machine check handler. This only handles real - * exceptions when something got corrupted coming in through int 18. + * 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 + * This is executed in #MC 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! * * On Intel systems this is entered on all CPUs in parallel through @@ -1343,6 +1349,14 @@ static void queue_task_work(struct mce * * issues: if the machine check was due to a failure of the memory * backing the user stack, tracing that reads the user stack will cause * potentially infinite recursion. + * + * Currently, the #MC handler calls out to a number of external facilities + * and, therefore, allows instrumentation around them. The optimal thing to + * have would be to do the absolutely minimal work required in #MC context + * and have instrumentation disabled only around that. Further processing can + * then happen in process context where instrumentation is allowed. Achieving + * that requires careful auditing and modifications. Until then, the code + * allows instrumentation temporarily, where required. * */ noinstr void do_machine_check(struct pt_regs *regs) {