| Message ID | 20170214020143.29713-1-ying.huang@intel.com (mailing list archive) |
|---|---|
| State | Changes Requested, archived |
| Headers | show |
On 2/13/2017 7:01 PM, Huang, Ying wrote: > From: Huang Ying <ying.huang@intel.com> > > It was reported that some firmware will use ACPI NVS area for BERT > address range. This will cause resources conflict because the ACPI > NVS area is marked as busy already. Fix this via excluding ACPI NVS > area when requesting IO resources for BERT. > > Reported-and-tested-by: Hans Kristian Rosbach <hansr@raskesider.no> > Signed-off-by: "Huang, Ying" <ying.huang@intel.com> Tested-by: Tyler Baicar <tbaicar@codeaurora.org>
On Tuesday, February 14, 2017 10:01:13 AM Huang, Ying wrote: > From: Huang Ying <ying.huang@intel.com> > > It was reported that some firmware will use ACPI NVS area for BERT > address range. This will cause resources conflict because the ACPI > NVS area is marked as busy already. Fix this via excluding ACPI NVS > area when requesting IO resources for BERT. > > Reported-and-tested-by: Hans Kristian Rosbach <hansr@raskesider.no> > Signed-off-by: "Huang, Ying" <ying.huang@intel.com> Boris, what do you think? > --- > drivers/acpi/apei/bert.c | 20 ++++++++++++-------- > 1 file changed, 12 insertions(+), 8 deletions(-) > > diff --git a/drivers/acpi/apei/bert.c b/drivers/acpi/apei/bert.c > index a05b5c0cf181..12771fcf0417 100644 > --- a/drivers/acpi/apei/bert.c > +++ b/drivers/acpi/apei/bert.c > @@ -97,6 +97,7 @@ static int __init bert_check_table(struct acpi_table_bert *bert_tab) > > static int __init bert_init(void) > { > + struct apei_resources bert_resources; > struct acpi_bert_region *boot_error_region; > struct acpi_table_bert *bert_tab; > unsigned int region_len; > @@ -127,13 +128,14 @@ static int __init bert_init(void) > } > > region_len = bert_tab->region_length; > - if (!request_mem_region(bert_tab->address, region_len, "APEI BERT")) { > - pr_err("Can't request iomem region <%016llx-%016llx>.\n", > - (unsigned long long)bert_tab->address, > - (unsigned long long)bert_tab->address + region_len - 1); > - return -EIO; > - } > - > + apei_resources_init(&bert_resources); > + rc = apei_resources_add(&bert_resources, bert_tab->address, > + region_len, true); > + if (rc) > + return rc; > + rc = apei_resources_request(&bert_resources, "APEI BERT"); > + if (rc) > + goto out_fini; > boot_error_region = ioremap_cache(bert_tab->address, region_len); > if (boot_error_region) { > bert_print_all(boot_error_region, region_len); > @@ -142,7 +144,9 @@ static int __init bert_init(void) > rc = -ENOMEM; > } > > - release_mem_region(bert_tab->address, region_len); > + apei_resources_release(&bert_resources); > +out_fini: > + apei_resources_fini(&bert_resources); > > return rc; > } > -- To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Thu, Feb 16, 2017 at 12:42:00AM +0100, Rafael J. Wysocki wrote: > On Tuesday, February 14, 2017 10:01:13 AM Huang, Ying wrote: > > From: Huang Ying <ying.huang@intel.com> > > > > It was reported that some firmware will use ACPI NVS area for BERT > > address range. This will cause resources conflict because the ACPI > > NVS area is marked as busy already. Fix this via excluding ACPI NVS > > area when requesting IO resources for BERT. > > > > Reported-and-tested-by: Hans Kristian Rosbach <hansr@raskesider.no> > > Signed-off-by: "Huang, Ying" <ying.huang@intel.com> > > Boris, what do you think? Lemme see, so the BERT is for hw errors which have happened during the previous boot and the machine couldn't handle them. So they do get saved in some mem in the fw for inspection during the next boot. And "some firmware" has decided to write them into non-volatile storage - because you should never ever forget those errors! :-) And I don't understand the "fix" here: we're excluding the NVS area when mapping the BERT table but then how are we supposed to read those errors from there? Or am I missing something obvious?
Borislav Petkov <bp@suse.de> writes: > On Thu, Feb 16, 2017 at 12:42:00AM +0100, Rafael J. Wysocki wrote: >> On Tuesday, February 14, 2017 10:01:13 AM Huang, Ying wrote: >> > From: Huang Ying <ying.huang@intel.com> >> > >> > It was reported that some firmware will use ACPI NVS area for BERT >> > address range. This will cause resources conflict because the ACPI >> > NVS area is marked as busy already. Fix this via excluding ACPI NVS >> > area when requesting IO resources for BERT. >> > >> > Reported-and-tested-by: Hans Kristian Rosbach <hansr@raskesider.no> >> > Signed-off-by: "Huang, Ying" <ying.huang@intel.com> >> >> Boris, what do you think? > > Lemme see, so the BERT is for hw errors which have happened during the > previous boot and the machine couldn't handle them. So they do get saved > in some mem in the fw for inspection during the next boot. > > And "some firmware" has decided to write them into non-volatile storage > - because you should never ever forget those errors! :-) > > And I don't understand the "fix" here: we're excluding the NVS area when > mapping the BERT table but then how are we supposed to read those errors > from there? The NVS area is excluded when request the resources, because the NVS area has been marked as busy already. But the whole BERT memory area is mapped, so we can read from it. > Or am I missing something obvious? Best Regards, Huang, Ying -- To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Thu, Feb 16, 2017 at 06:24:39PM +0800, Huang, Ying wrote: > The NVS area is excluded when request the resources, because the NVS > area has been marked as busy already. But the whole BERT memory area is > mapped, so we can read from it. So you're saying the NVS area is part of the BERT area and we can access the whole range?
Borislav Petkov <bp@suse.de> writes: > On Thu, Feb 16, 2017 at 06:24:39PM +0800, Huang, Ying wrote: >> The NVS area is excluded when request the resources, because the NVS >> area has been marked as busy already. But the whole BERT memory area is >> mapped, so we can read from it. > > So you're saying the NVS area is part of the BERT area and we can access > the whole range? I am not sure whether the NVS area is a part of the BERT area, but apparently they are overlapped in some way. We will access the whole BERT area here via ioremap the whole range. Best Regards, Huang, Ying -- To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, Feb 17, 2017 at 08:31:12AM +0800, Huang, Ying wrote: > I am not sure whether the NVS area is a part of the BERT area, but > apparently they are overlapped in some way. We will access the whole > BERT area here via ioremap the whole range. This is the part that was missing from the commit message - explaining *why* we need this fix.
Borislav Petkov <bp@suse.de> writes: > On Fri, Feb 17, 2017 at 08:31:12AM +0800, Huang, Ying wrote: >> I am not sure whether the NVS area is a part of the BERT area, but >> apparently they are overlapped in some way. We will access the whole >> BERT area here via ioremap the whole range. > > This is the part that was missing from the commit message - explaining > *why* we need this fix. OK. I will revise the patch description accordingly. Best Regards, Huang, Ying -- To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/acpi/apei/bert.c b/drivers/acpi/apei/bert.c index a05b5c0cf181..12771fcf0417 100644 --- a/drivers/acpi/apei/bert.c +++ b/drivers/acpi/apei/bert.c @@ -97,6 +97,7 @@ static int __init bert_check_table(struct acpi_table_bert *bert_tab) static int __init bert_init(void) { + struct apei_resources bert_resources; struct acpi_bert_region *boot_error_region; struct acpi_table_bert *bert_tab; unsigned int region_len; @@ -127,13 +128,14 @@ static int __init bert_init(void) } region_len = bert_tab->region_length; - if (!request_mem_region(bert_tab->address, region_len, "APEI BERT")) { - pr_err("Can't request iomem region <%016llx-%016llx>.\n", - (unsigned long long)bert_tab->address, - (unsigned long long)bert_tab->address + region_len - 1); - return -EIO; - } - + apei_resources_init(&bert_resources); + rc = apei_resources_add(&bert_resources, bert_tab->address, + region_len, true); + if (rc) + return rc; + rc = apei_resources_request(&bert_resources, "APEI BERT"); + if (rc) + goto out_fini; boot_error_region = ioremap_cache(bert_tab->address, region_len); if (boot_error_region) { bert_print_all(boot_error_region, region_len); @@ -142,7 +144,9 @@ static int __init bert_init(void) rc = -ENOMEM; } - release_mem_region(bert_tab->address, region_len); + apei_resources_release(&bert_resources); +out_fini: + apei_resources_fini(&bert_resources); return rc; }