diff options
| -rw-r--r-- | drivers/acpi/acpica/hwregs.c | 33 |
1 files changed, 24 insertions, 9 deletions
diff --git a/drivers/acpi/acpica/hwregs.c b/drivers/acpi/acpica/hwregs.c index 9c8162128c2b..5a64e577975f 100644 --- a/drivers/acpi/acpica/hwregs.c +++ b/drivers/acpi/acpica/hwregs.c @@ -372,9 +372,17 @@ acpi_hw_read_multiple(u32 *value, } } - /* Shift the B bits above the A bits */ - - *value = value_a | (value_b << register_a->bit_width); + /* + * OR the two return values together. No shifting or masking is necessary, + * because of how the PM1 registers are defined in the ACPI specification: + * + * "Although the bits can be split between the two register blocks (each + * register block has a unique pointer within the FADT), the bit positions + * are maintained. The register block with unimplemented bits (that is, + * those implemented in the other register block) always returns zeros, + * and writes have no side effects" + */ + *value = (value_a | value_b); return (AE_OK); } @@ -406,13 +414,20 @@ acpi_hw_write_multiple(u32 value, return (status); } - /* Second register is optional */ - + /* + * Second register is optional + * + * No bit shifting or clearing is necessary, because of how the PM1 + * registers are defined in the ACPI specification: + * + * "Although the bits can be split between the two register blocks (each + * register block has a unique pointer within the FADT), the bit positions + * are maintained. The register block with unimplemented bits (that is, + * those implemented in the other register block) always returns zeros, + * and writes have no side effects" + */ if (register_b->address) { - - /* Normalize the B bits before write */ - - status = acpi_write(value >> register_a->bit_width, register_b); + status = acpi_write(value, register_b); } return (status); |