]> bbs.cooldavid.org Git - net-next-2.6.git/commitdiff
ACPI: create "processor.bm_check_disable" boot param
authorLen Brown <len.brown@intel.com>
Thu, 22 Jul 2010 21:23:10 +0000 (17:23 -0400)
committerLen Brown <len.brown@intel.com>
Thu, 22 Jul 2010 21:23:10 +0000 (17:23 -0400)
processor.bm_check_disable=1" prevents Linux from checking BM_STS
before entering C3-type cpu power states.

This may be useful for a system running acpi_idle
where the BIOS exports FADT C-states, _CST IO C-states,
or _CST FFH C-states with the BM_STS bit set;
while configuring the chipset to set BM_STS
more frequently than perhaps is optimal.

Note that such systems may have been developed
using a tickful OS that would quickly clear BM_STS,
rather than a tickless OS that may go for some time
between checking and clearing BM_STS.

Note also that an alternative for newer systems
is to use the intel_idle driver, which always
ignores BM_STS, relying Linux device drivers
to register constraints explicitly via PM_QOS.

https://bugzilla.kernel.org/show_bug.cgi?id=15886

Signed-off-by: Len Brown <len.brown@intel.com>
drivers/acpi/processor_idle.c

index b351342f1faf636a1556414e061f6927f9c71d95..1d410485529628969a4a66e58b441540b4956b62 100644 (file)
@@ -76,6 +76,8 @@ static unsigned int max_cstate __read_mostly = ACPI_PROCESSOR_MAX_POWER;
 module_param(max_cstate, uint, 0000);
 static unsigned int nocst __read_mostly;
 module_param(nocst, uint, 0000);
+static int bm_check_disable __read_mostly;
+module_param(bm_check_disable, uint, 0000);
 
 static unsigned int latency_factor __read_mostly = 2;
 module_param(latency_factor, uint, 0644);
@@ -763,6 +765,9 @@ static int acpi_idle_bm_check(void)
 {
        u32 bm_status = 0;
 
+       if (bm_check_disable)
+               return 0;
+
        acpi_read_bit_register(ACPI_BITREG_BUS_MASTER_STATUS, &bm_status);
        if (bm_status)
                acpi_write_bit_register(ACPI_BITREG_BUS_MASTER_STATUS, 1);