In the ACPI 5.1 version of the spec, the struct for the GICC subtable (struct acpi_madt_generic_interrupt) of the MADT is 76 bytes long; in ACPI 6.0, the struct is 80 bytes long. But, there is only one definition in ACPICA for this struct -- and that is the 6.0 version. Hence, when BAD_MADT_ENTRY() compares the struct size to the length in the GICC subtable, it fails if 5.1 structs are in use, and there are systems in the wild that have them.
Note that this was found in linux-next and these patches apply against that tree and the arm64 kernel tree; 4.1 does not appear to have this problem since it still has the 5.1 struct definition.
Though there is precedent in ia64 code for ignoring the changes in size, this patch set instead verifies correctness. The first patch adds the BAD_MADT_GICC_ENTRY() macro to check the GICC subtable only, accounting for the difference in specification versions that are possible. The second patch replaces BAD_MADT_ENTRY usage with the BAD_MADT_GICC_ENTRY macro in arm64 code, which is currently the only architecture affected. The BAD_MADT_ENTRY() will continue to work as is for all other MADT subtables.
I have tested these patches on an APM Mustang with version 1.15 firmware, where the problem was found, and they fix the problem -- i.e., the system will boot with either Linux 4.1 or linux-next kernels using the same ACPI 5.1 compatible firmware.
Changes for v4: -- Reword the cover letter to reflect smaller patch set -- Simplify BAD_MADT_GICC_ENTRY to the minimum needed; this removed the need for the first patch containing version number macros (Rafael) -- Simplify determining the GICC subtable length (Catalin)
Changes for v3: -- Modified the macros for using spec version numbers in order to make them clearer (Rafael, Hanjun) -- Moved the definition of the BAD_MADT_GICC_ENTRY macro to an arm64-specific header file since only this architecture uses the GICC subtable (Rafael) -- Added Reviewed-by (Hanjun) and Acked-by (Will) tags to 3/3, the only unchanged patch; other tags could be applied but the patches have changed. -- Added Fixes: tag to patches
Changes for v2: -- Replace magic constants with proper defines (Lorenzo) -- Minor syntax clean-up noted by checkpatch -- Send out CCs properly this time -- Minor clean-up of the paragraphs in this cover letter
Al Stone (2): ACPI / ARM64: add BAD_MADT_GICC_ENTRY() macro ACPI / ARM64 : use the new BAD_MADT_GICC_ENTRY macro
arch/arm64/include/asm/acpi.h | 8 ++++++++ arch/arm64/kernel/smp.c | 2 +- drivers/irqchip/irq-gic.c | 2 +- 3 files changed, 10 insertions(+), 2 deletions(-)
From: Al Stone <al.stone@.linaro.org>
The BAD_MADT_ENTRY() macro is designed to work for all of the subtables of the MADT. In the ACPI 5.1 version of the spec, the struct for the GICC subtable (struct acpi_madt_generic_interrupt) is 76 bytes long; in ACPI 6.0, the struct is 80 bytes long. But, there is only one definition in ACPICA for this struct -- and that is the 6.0 version. Hence, when BAD_MADT_ENTRY() compares the struct size to the length in the GICC subtable, it fails if 5.1 structs are in use, and there are systems in the wild that have them.
This patch adds the BAD_MADT_GICC_ENTRY() that checks the GICC subtable only, accounting for the difference in specification versions that are possible. The BAD_MADT_ENTRY() will continue to work as is for all other MADT subtables.
This code is being added to an arm64 header file since that is currently the only architecture using the GICC subtable of the MADT. As a GIC is specific to ARM, it is also unlikely the subtable will be used elsewhere.
Fixes: aeb823bbacc2 (ACPICA: ACPI 6.0: Add changes for FADT table.) Signed-off-by: Al Stone al.stone@linaro.org --- arch/arm64/include/asm/acpi.h | 8 ++++++++ 1 file changed, 8 insertions(+)
diff --git a/arch/arm64/include/asm/acpi.h b/arch/arm64/include/asm/acpi.h index 39248d3..c650c91 100644 --- a/arch/arm64/include/asm/acpi.h +++ b/arch/arm64/include/asm/acpi.h @@ -19,6 +19,14 @@ #include <asm/psci.h> #include <asm/smp_plat.h>
+/* Macros for consistency checks of the GICC subtable of MADT */ +#define ACPI_MADT_GICC_LENGTH \ + (acpi_gbl_FADT.header.revision < 6 ? 76 : 80) + +#define BAD_MADT_GICC_ENTRY(entry, end) ( \ + (!entry) || (unsigned long)entry + sizeof(*entry) > end || \ + entry->header.length != ACPI_MADT_GICC_LENGTH) + /* Basic configuration for ACPI */ #ifdef CONFIG_ACPI /* ACPI table mapping after acpi_gbl_permanent_mmap is set */
On Tue, Jul 07, 2015 at 12:16:47AM +0100, Al Stone wrote:
From: Al Stone <al.stone@.linaro.org>
The BAD_MADT_ENTRY() macro is designed to work for all of the subtables of the MADT. In the ACPI 5.1 version of the spec, the struct for the GICC subtable (struct acpi_madt_generic_interrupt) is 76 bytes long; in ACPI 6.0, the struct is 80 bytes long. But, there is only one definition in ACPICA for this struct -- and that is the 6.0 version. Hence, when BAD_MADT_ENTRY() compares the struct size to the length in the GICC subtable, it fails if 5.1 structs are in use, and there are systems in the wild that have them.
This patch adds the BAD_MADT_GICC_ENTRY() that checks the GICC subtable only, accounting for the difference in specification versions that are possible. The BAD_MADT_ENTRY() will continue to work as is for all other MADT subtables.
This code is being added to an arm64 header file since that is currently the only architecture using the GICC subtable of the MADT. As a GIC is specific to ARM, it is also unlikely the subtable will be used elsewhere.
Fixes: aeb823bbacc2 (ACPICA: ACPI 6.0: Add changes for FADT table.) Signed-off-by: Al Stone al.stone@linaro.org
arch/arm64/include/asm/acpi.h | 8 ++++++++ 1 file changed, 8 insertions(+)
Not the nicest patch I've ever seen, but if it gets things working again:
Acked-by: Will Deacon will.deacon@arm.com
Catalin, I assume you're picking these two up for 4.2?
Will
On 07/07/2015 03:25 AM, Will Deacon wrote:
On Tue, Jul 07, 2015 at 12:16:47AM +0100, Al Stone wrote:
From: Al Stone <al.stone@.linaro.org>
The BAD_MADT_ENTRY() macro is designed to work for all of the subtables of the MADT. In the ACPI 5.1 version of the spec, the struct for the GICC subtable (struct acpi_madt_generic_interrupt) is 76 bytes long; in ACPI 6.0, the struct is 80 bytes long. But, there is only one definition in ACPICA for this struct -- and that is the 6.0 version. Hence, when BAD_MADT_ENTRY() compares the struct size to the length in the GICC subtable, it fails if 5.1 structs are in use, and there are systems in the wild that have them.
This patch adds the BAD_MADT_GICC_ENTRY() that checks the GICC subtable only, accounting for the difference in specification versions that are possible. The BAD_MADT_ENTRY() will continue to work as is for all other MADT subtables.
This code is being added to an arm64 header file since that is currently the only architecture using the GICC subtable of the MADT. As a GIC is specific to ARM, it is also unlikely the subtable will be used elsewhere.
Fixes: aeb823bbacc2 (ACPICA: ACPI 6.0: Add changes for FADT table.) Signed-off-by: Al Stone al.stone@linaro.org
arch/arm64/include/asm/acpi.h | 8 ++++++++ 1 file changed, 8 insertions(+)
Not the nicest patch I've ever seen, but if it gets things working again:
Acked-by: Will Deacon will.deacon@arm.com
Catalin, I assume you're picking these two up for 4.2?
Will
Yeah, not my favorite either, but it does work. This will get cleaned up by fixing the larger problems, I believe, and then we should be able to remove this one. That's the plan, at any rate.
For those parts of the arm64 ACPI code that need to check GICC subtables in the MADT, use the new BAD_MADT_GICC_ENTRY macro instead of the previous BAD_MADT_ENTRY. The new macro takes into account differences in the size of the GICC subtable that the old macro did not; this caused failures even though the subtable entries are valid.
Fixes: aeb823bbacc2 (ACPICA: ACPI 6.0: Add changes for FADT table.) Signed-off-by: Al Stone al.stone@linaro.org Reviewed-by: Hanjun Guo hanjun.guo@linaro.org Acked-by: Will Deacon will.deacon@arm.com --- arch/arm64/kernel/smp.c | 2 +- drivers/irqchip/irq-gic.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/arch/arm64/kernel/smp.c b/arch/arm64/kernel/smp.c index a1883bf..25fc88c 100644 --- a/arch/arm64/kernel/smp.c +++ b/arch/arm64/kernel/smp.c @@ -438,7 +438,7 @@ acpi_parse_gic_cpu_interface(struct acpi_subtable_header *header, struct acpi_madt_generic_interrupt *processor;
processor = (struct acpi_madt_generic_interrupt *)header; - if (BAD_MADT_ENTRY(processor, end)) + if (BAD_MADT_GICC_ENTRY(processor, end)) return -EINVAL;
acpi_table_print_madt_entry(header); diff --git a/drivers/irqchip/irq-gic.c b/drivers/irqchip/irq-gic.c index 8d7e1c8..4dd8826 100644 --- a/drivers/irqchip/irq-gic.c +++ b/drivers/irqchip/irq-gic.c @@ -1055,7 +1055,7 @@ gic_acpi_parse_madt_cpu(struct acpi_subtable_header *header,
processor = (struct acpi_madt_generic_interrupt *)header;
- if (BAD_MADT_ENTRY(processor, end)) + if (BAD_MADT_GICC_ENTRY(processor, end)) return -EINVAL;
/*
Hi Al,
On Tue, Jul 7, 2015 at 1:16 AM, Al Stone al.stone@linaro.org wrote:
In the ACPI 5.1 version of the spec, the struct for the GICC subtable (struct acpi_madt_generic_interrupt) of the MADT is 76 bytes long; in ACPI 6.0, the struct is 80 bytes long. But, there is only one definition in ACPICA for this struct -- and that is the 6.0 version. Hence, when BAD_MADT_ENTRY() compares the struct size to the length in the GICC subtable, it fails if 5.1 structs are in use, and there are systems in the wild that have them.
Note that this was found in linux-next and these patches apply against that tree and the arm64 kernel tree; 4.1 does not appear to have this problem since it still has the 5.1 struct definition.
Though there is precedent in ia64 code for ignoring the changes in size, this patch set instead verifies correctness. The first patch adds the BAD_MADT_GICC_ENTRY() macro to check the GICC subtable only, accounting for the difference in specification versions that are possible. The second patch replaces BAD_MADT_ENTRY usage with the BAD_MADT_GICC_ENTRY macro in arm64 code, which is currently the only architecture affected. The BAD_MADT_ENTRY() will continue to work as is for all other MADT subtables.
I have tested these patches on an APM Mustang with version 1.15 firmware, where the problem was found, and they fix the problem -- i.e., the system will boot with either Linux 4.1 or linux-next kernels using the same ACPI 5.1 compatible firmware.
ACK for the series, but I guess it's better to let it go via ARM64, right?
Rafael
On 07/06/2015 05:20 PM, Rafael J. Wysocki wrote:
Hi Al,
On Tue, Jul 7, 2015 at 1:16 AM, Al Stone al.stone@linaro.org wrote:
In the ACPI 5.1 version of the spec, the struct for the GICC subtable (struct acpi_madt_generic_interrupt) of the MADT is 76 bytes long; in ACPI 6.0, the struct is 80 bytes long. But, there is only one definition in ACPICA for this struct -- and that is the 6.0 version. Hence, when BAD_MADT_ENTRY() compares the struct size to the length in the GICC subtable, it fails if 5.1 structs are in use, and there are systems in the wild that have them.
Note that this was found in linux-next and these patches apply against that tree and the arm64 kernel tree; 4.1 does not appear to have this problem since it still has the 5.1 struct definition.
Though there is precedent in ia64 code for ignoring the changes in size, this patch set instead verifies correctness. The first patch adds the BAD_MADT_GICC_ENTRY() macro to check the GICC subtable only, accounting for the difference in specification versions that are possible. The second patch replaces BAD_MADT_ENTRY usage with the BAD_MADT_GICC_ENTRY macro in arm64 code, which is currently the only architecture affected. The BAD_MADT_ENTRY() will continue to work as is for all other MADT subtables.
I have tested these patches on an APM Mustang with version 1.15 firmware, where the problem was found, and they fix the problem -- i.e., the system will boot with either Linux 4.1 or linux-next kernels using the same ACPI 5.1 compatible firmware.
ACK for the series, but I guess it's better to let it go via ARM64, right?
Rafael
Thanks, Rafael. Yeah, probably so. Will has ACKd the one patch (2/2); if he and/or Catalin ACK patch 1/2, then this seems like it would pretty cleanly fit into ARM64. The only question would be if Will or Catalin would want an ACK from Thomas on the irq-gic.c part in 2/2.
On Mon, 6 Jul 2015, Al Stone wrote:
On 07/06/2015 05:20 PM, Rafael J. Wysocki wrote:
Hi Al,
On Tue, Jul 7, 2015 at 1:16 AM, Al Stone al.stone@linaro.org wrote:
In the ACPI 5.1 version of the spec, the struct for the GICC subtable (struct acpi_madt_generic_interrupt) of the MADT is 76 bytes long; in ACPI 6.0, the struct is 80 bytes long. But, there is only one definition in ACPICA for this struct -- and that is the 6.0 version. Hence, when BAD_MADT_ENTRY() compares the struct size to the length in the GICC subtable, it fails if 5.1 structs are in use, and there are systems in the wild that have them.
Note that this was found in linux-next and these patches apply against that tree and the arm64 kernel tree; 4.1 does not appear to have this problem since it still has the 5.1 struct definition.
Though there is precedent in ia64 code for ignoring the changes in size, this patch set instead verifies correctness. The first patch adds the BAD_MADT_GICC_ENTRY() macro to check the GICC subtable only, accounting for the difference in specification versions that are possible. The second patch replaces BAD_MADT_ENTRY usage with the BAD_MADT_GICC_ENTRY macro in arm64 code, which is currently the only architecture affected. The BAD_MADT_ENTRY() will continue to work as is for all other MADT subtables.
I have tested these patches on an APM Mustang with version 1.15 firmware, where the problem was found, and they fix the problem -- i.e., the system will boot with either Linux 4.1 or linux-next kernels using the same ACPI 5.1 compatible firmware.
ACK for the series, but I guess it's better to let it go via ARM64, right?
Rafael
Thanks, Rafael. Yeah, probably so. Will has ACKd the one patch (2/2); if he and/or Catalin ACK patch 1/2, then this seems like it would pretty cleanly fit into ARM64. The only question would be if Will or Catalin would want an ACK from Thomas on the irq-gic.c part in 2/2.
No objections from my side.
Thanks,
tglx
On Tue, Jul 07, 2015 at 01:20:51AM +0200, Rafael J. Wysocki wrote:
On Tue, Jul 7, 2015 at 1:16 AM, Al Stone al.stone@linaro.org wrote:
In the ACPI 5.1 version of the spec, the struct for the GICC subtable (struct acpi_madt_generic_interrupt) of the MADT is 76 bytes long; in ACPI 6.0, the struct is 80 bytes long. But, there is only one definition in ACPICA for this struct -- and that is the 6.0 version. Hence, when BAD_MADT_ENTRY() compares the struct size to the length in the GICC subtable, it fails if 5.1 structs are in use, and there are systems in the wild that have them.
Note that this was found in linux-next and these patches apply against that tree and the arm64 kernel tree; 4.1 does not appear to have this problem since it still has the 5.1 struct definition.
Though there is precedent in ia64 code for ignoring the changes in size, this patch set instead verifies correctness. The first patch adds the BAD_MADT_GICC_ENTRY() macro to check the GICC subtable only, accounting for the difference in specification versions that are possible. The second patch replaces BAD_MADT_ENTRY usage with the BAD_MADT_GICC_ENTRY macro in arm64 code, which is currently the only architecture affected. The BAD_MADT_ENTRY() will continue to work as is for all other MADT subtables.
I have tested these patches on an APM Mustang with version 1.15 firmware, where the problem was found, and they fix the problem -- i.e., the system will boot with either Linux 4.1 or linux-next kernels using the same ACPI 5.1 compatible firmware.
ACK for the series, but I guess it's better to let it go via ARM64, right?
Fine by me. I'll pick them up for 4.2-rc2.
Thanks.
On 07/07/2015 07:31 AM, Catalin Marinas wrote:
On Tue, Jul 07, 2015 at 01:20:51AM +0200, Rafael J. Wysocki wrote:
On Tue, Jul 7, 2015 at 1:16 AM, Al Stone al.stone@linaro.org wrote:
In the ACPI 5.1 version of the spec, the struct for the GICC subtable (struct acpi_madt_generic_interrupt) of the MADT is 76 bytes long; in ACPI 6.0, the struct is 80 bytes long. But, there is only one definition in ACPICA for this struct -- and that is the 6.0 version. Hence, when BAD_MADT_ENTRY() compares the struct size to the length in the GICC subtable, it fails if 5.1 structs are in use, and there are systems in the wild that have them.
Note that this was found in linux-next and these patches apply against that tree and the arm64 kernel tree; 4.1 does not appear to have this problem since it still has the 5.1 struct definition.
Though there is precedent in ia64 code for ignoring the changes in size, this patch set instead verifies correctness. The first patch adds the BAD_MADT_GICC_ENTRY() macro to check the GICC subtable only, accounting for the difference in specification versions that are possible. The second patch replaces BAD_MADT_ENTRY usage with the BAD_MADT_GICC_ENTRY macro in arm64 code, which is currently the only architecture affected. The BAD_MADT_ENTRY() will continue to work as is for all other MADT subtables.
I have tested these patches on an APM Mustang with version 1.15 firmware, where the problem was found, and they fix the problem -- i.e., the system will boot with either Linux 4.1 or linux-next kernels using the same ACPI 5.1 compatible firmware.
ACK for the series, but I guess it's better to let it go via ARM64, right?
Fine by me. I'll pick them up for 4.2-rc2.
Thanks.
Thanks, Catalin. Holler if there's any problems.
-
Al Stone -
Al Stone -
Catalin Marinas -
Rafael J. Wysocki -
Thomas Gleixner -
Will Deacon