| Message ID | 1516969885-150532-3-git-send-email-xiexiuqi@huawei.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Fri, Jan 26, 2018 at 08:31:24PM +0800, Xie XiuQi wrote: > Add a notify chain for process memory section, with > which other modules might do error recovery. > > Signed-off-by: Xie XiuQi <xiexiuqi@huawei.com> > Tested-by: Wang Xiongfeng <wangxiongfeng2@huawei.com> > Tested-by: Tyler Baicar <tbaicar@codeaurora.org> > --- > drivers/acpi/apei/ghes.c | 10 ++++++++++ > include/acpi/ghes.h | 8 ++++++++ > 2 files changed, 18 insertions(+) > > diff --git a/drivers/acpi/apei/ghes.c b/drivers/acpi/apei/ghes.c > index cff671d..1f0ebfb 100644 > --- a/drivers/acpi/apei/ghes.c > +++ b/drivers/acpi/apei/ghes.c > @@ -109,6 +109,9 @@ static inline bool is_hest_type_generic_v2(struct ghes *ghes) > static LIST_HEAD(ghes_hed); > static DEFINE_MUTEX(ghes_list_mutex); > > +ATOMIC_NOTIFIER_HEAD(ghes_mem_err_chain); > +EXPORT_SYMBOL(ghes_mem_err_chain); Definitely not EXPORT_SYMBOL. And certainly not export the notifier head. Have register and unregister functions instead. That is, *if* you can't solve it differently with James. Notifiers should be avoided if possible.
Hi Boris, Thanks for your comments. On 2018/2/7 18:31, Borislav Petkov wrote: > On Fri, Jan 26, 2018 at 08:31:24PM +0800, Xie XiuQi wrote: >> Add a notify chain for process memory section, with >> which other modules might do error recovery. >> >> Signed-off-by: Xie XiuQi <xiexiuqi@huawei.com> >> Tested-by: Wang Xiongfeng <wangxiongfeng2@huawei.com> >> Tested-by: Tyler Baicar <tbaicar@codeaurora.org> >> --- >> drivers/acpi/apei/ghes.c | 10 ++++++++++ >> include/acpi/ghes.h | 8 ++++++++ >> 2 files changed, 18 insertions(+) >> >> diff --git a/drivers/acpi/apei/ghes.c b/drivers/acpi/apei/ghes.c >> index cff671d..1f0ebfb 100644 >> --- a/drivers/acpi/apei/ghes.c >> +++ b/drivers/acpi/apei/ghes.c >> @@ -109,6 +109,9 @@ static inline bool is_hest_type_generic_v2(struct ghes *ghes) >> static LIST_HEAD(ghes_hed); >> static DEFINE_MUTEX(ghes_list_mutex); >> >> +ATOMIC_NOTIFIER_HEAD(ghes_mem_err_chain); >> +EXPORT_SYMBOL(ghes_mem_err_chain); > > Definitely not EXPORT_SYMBOL. > > And certainly not export the notifier head. Have register and unregister > functions instead. > > That is, *if* you can't solve it differently with James. Notifiers > should be avoided if possible. > Yes, I saw the patchset, I agree. I will work with James to solve the problem.
diff --git a/drivers/acpi/apei/ghes.c b/drivers/acpi/apei/ghes.c index cff671d..1f0ebfb 100644 --- a/drivers/acpi/apei/ghes.c +++ b/drivers/acpi/apei/ghes.c @@ -109,6 +109,9 @@ static inline bool is_hest_type_generic_v2(struct ghes *ghes) static LIST_HEAD(ghes_hed); static DEFINE_MUTEX(ghes_list_mutex); +ATOMIC_NOTIFIER_HEAD(ghes_mem_err_chain); +EXPORT_SYMBOL(ghes_mem_err_chain); + /* * Because the memory area used to transfer hardware error information * from BIOS to Linux can be determined only in NMI, IRQ or timer @@ -441,6 +444,13 @@ static void ghes_do_proc(struct ghes *ghes, if (guid_equal(sec_type, &CPER_SEC_PLATFORM_MEM)) { struct cper_sec_mem_err *mem_err = acpi_hest_get_payload(gdata); + struct ghes_mem_err mem; + + mem.notify_type = ghes->generic->notify.type; + mem.severity = gdata->error_severity; + mem.mem_err = mem_err; + + atomic_notifier_call_chain(&ghes_mem_err_chain, 0, &mem); ghes_edac_report_mem_error(ghes, sev, mem_err); diff --git a/include/acpi/ghes.h b/include/acpi/ghes.h index d626367..921ebf1 100644 --- a/include/acpi/ghes.h +++ b/include/acpi/ghes.h @@ -124,4 +124,12 @@ static inline void *acpi_hest_get_next(struct acpi_hest_generic_data *gdata) extern void ghes_arm_process_error(struct ghes *ghes, struct cper_sec_proc_arm *err); +struct ghes_mem_err { + int notify_type; + int severity; + struct cper_sec_mem_err *mem_err; +}; + +extern struct atomic_notifier_head ghes_mem_err_chain; + #endif /* GHES_H */