aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
-rw-r--r--include/linux/mm.h1
-rw-r--r--mm/memory-failure.c50
2 files changed, 30 insertions, 21 deletions
diff --git a/include/linux/mm.h b/include/linux/mm.h
index bcc523474724..bf169ca69812 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1606,6 +1606,7 @@ void vmemmap_populate_print_last(void);
1606 1606
1607enum mf_flags { 1607enum mf_flags {
1608 MF_COUNT_INCREASED = 1 << 0, 1608 MF_COUNT_INCREASED = 1 << 0,
1609 MF_ACTION_REQUIRED = 1 << 1,
1609}; 1610};
1610extern int memory_failure(unsigned long pfn, int trapno, int flags); 1611extern int memory_failure(unsigned long pfn, int trapno, int flags);
1611extern void memory_failure_queue(unsigned long pfn, int trapno, int flags); 1612extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
diff --git a/mm/memory-failure.c b/mm/memory-failure.c
index ab259bb0adc5..95fd307ebb30 100644
--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -187,33 +187,40 @@ int hwpoison_filter(struct page *p)
187EXPORT_SYMBOL_GPL(hwpoison_filter); 187EXPORT_SYMBOL_GPL(hwpoison_filter);
188 188
189/* 189/*
190 * Send all the processes who have the page mapped an ``action optional'' 190 * Send all the processes who have the page mapped a signal.
191 * signal. 191 * ``action optional'' if they are not immediately affected by the error
192 * ``action required'' if error happened in current execution context
192 */ 193 */
193static int kill_proc_ao(struct task_struct *t, unsigned long addr, int trapno, 194static int kill_proc(struct task_struct *t, unsigned long addr, int trapno,
194 unsigned long pfn, struct page *page) 195 unsigned long pfn, struct page *page, int flags)
195{ 196{
196 struct siginfo si; 197 struct siginfo si;
197 int ret; 198 int ret;
198 199
199 printk(KERN_ERR 200 printk(KERN_ERR
200 "MCE %#lx: Killing %s:%d early due to hardware memory corruption\n", 201 "MCE %#lx: Killing %s:%d due to hardware memory corruption\n",
201 pfn, t->comm, t->pid); 202 pfn, t->comm, t->pid);
202 si.si_signo = SIGBUS; 203 si.si_signo = SIGBUS;
203 si.si_errno = 0; 204 si.si_errno = 0;
204 si.si_code = BUS_MCEERR_AO;
205 si.si_addr = (void *)addr; 205 si.si_addr = (void *)addr;
206#ifdef __ARCH_SI_TRAPNO 206#ifdef __ARCH_SI_TRAPNO
207 si.si_trapno = trapno; 207 si.si_trapno = trapno;
208#endif 208#endif
209 si.si_addr_lsb = compound_trans_order(compound_head(page)) + PAGE_SHIFT; 209 si.si_addr_lsb = compound_trans_order(compound_head(page)) + PAGE_SHIFT;
210 /* 210
211 * Don't use force here, it's convenient if the signal 211 if ((flags & MF_ACTION_REQUIRED) && t == current) {
212 * can be temporarily blocked. 212 si.si_code = BUS_MCEERR_AR;
213 * This could cause a loop when the user sets SIGBUS 213 ret = force_sig_info(SIGBUS, &si, t);
214 * to SIG_IGN, but hopefully no one will do that? 214 } else {
215 */ 215 /*
216 ret = send_sig_info(SIGBUS, &si, t); /* synchronous? */ 216 * Don't use force here, it's convenient if the signal
217 * can be temporarily blocked.
218 * This could cause a loop when the user sets SIGBUS
219 * to SIG_IGN, but hopefully no one will do that?
220 */
221 si.si_code = BUS_MCEERR_AO;
222 ret = send_sig_info(SIGBUS, &si, t); /* synchronous? */
223 }
217 if (ret < 0) 224 if (ret < 0)
218 printk(KERN_INFO "MCE: Error sending signal to %s:%d: %d\n", 225 printk(KERN_INFO "MCE: Error sending signal to %s:%d: %d\n",
219 t->comm, t->pid, ret); 226 t->comm, t->pid, ret);
@@ -338,8 +345,9 @@ static void add_to_kill(struct task_struct *tsk, struct page *p,
338 * Also when FAIL is set do a force kill because something went 345 * Also when FAIL is set do a force kill because something went
339 * wrong earlier. 346 * wrong earlier.
340 */ 347 */
341static void kill_procs_ao(struct list_head *to_kill, int doit, int trapno, 348static void kill_procs(struct list_head *to_kill, int doit, int trapno,
342 int fail, struct page *page, unsigned long pfn) 349 int fail, struct page *page, unsigned long pfn,
350 int flags)
343{ 351{
344 struct to_kill *tk, *next; 352 struct to_kill *tk, *next;
345 353
@@ -363,8 +371,8 @@ static void kill_procs_ao(struct list_head *to_kill, int doit, int trapno,
363 * check for that, but we need to tell the 371 * check for that, but we need to tell the
364 * process anyways. 372 * process anyways.
365 */ 373 */
366 else if (kill_proc_ao(tk->tsk, tk->addr, trapno, 374 else if (kill_proc(tk->tsk, tk->addr, trapno,
367 pfn, page) < 0) 375 pfn, page, flags) < 0)
368 printk(KERN_ERR 376 printk(KERN_ERR
369 "MCE %#lx: Cannot send advisory machine check signal to %s:%d\n", 377 "MCE %#lx: Cannot send advisory machine check signal to %s:%d\n",
370 pfn, tk->tsk->comm, tk->tsk->pid); 378 pfn, tk->tsk->comm, tk->tsk->pid);
@@ -844,7 +852,7 @@ static int page_action(struct page_state *ps, struct page *p,
844 * the pages and send SIGBUS to the processes if the data was dirty. 852 * the pages and send SIGBUS to the processes if the data was dirty.
845 */ 853 */
846static int hwpoison_user_mappings(struct page *p, unsigned long pfn, 854static int hwpoison_user_mappings(struct page *p, unsigned long pfn,
847 int trapno) 855 int trapno, int flags)
848{ 856{
849 enum ttu_flags ttu = TTU_UNMAP | TTU_IGNORE_MLOCK | TTU_IGNORE_ACCESS; 857 enum ttu_flags ttu = TTU_UNMAP | TTU_IGNORE_MLOCK | TTU_IGNORE_ACCESS;
850 struct address_space *mapping; 858 struct address_space *mapping;
@@ -962,8 +970,8 @@ static int hwpoison_user_mappings(struct page *p, unsigned long pfn,
962 * use a more force-full uncatchable kill to prevent 970 * use a more force-full uncatchable kill to prevent
963 * any accesses to the poisoned memory. 971 * any accesses to the poisoned memory.
964 */ 972 */
965 kill_procs_ao(&tokill, !!PageDirty(ppage), trapno, 973 kill_procs(&tokill, !!PageDirty(ppage), trapno,
966 ret != SWAP_SUCCESS, p, pfn); 974 ret != SWAP_SUCCESS, p, pfn, flags);
967 975
968 return ret; 976 return ret;
969} 977}
@@ -1148,7 +1156,7 @@ int memory_failure(unsigned long pfn, int trapno, int flags)
1148 * Now take care of user space mappings. 1156 * Now take care of user space mappings.
1149 * Abort on fail: __delete_from_page_cache() assumes unmapped page. 1157 * Abort on fail: __delete_from_page_cache() assumes unmapped page.
1150 */ 1158 */
1151 if (hwpoison_user_mappings(p, pfn, trapno) != SWAP_SUCCESS) { 1159 if (hwpoison_user_mappings(p, pfn, trapno, flags) != SWAP_SUCCESS) {
1152 printk(KERN_ERR "MCE %#lx: cannot unmap page, give up\n", pfn); 1160 printk(KERN_ERR "MCE %#lx: cannot unmap page, give up\n", pfn);
1153 res = -EBUSY; 1161 res = -EBUSY;
1154 goto out; 1162 goto out;