| Message ID | c66cf9c07985d4fd34738dae927abd3d0383896a.1421234254.git.lv.zheng@intel.com (mailing list archive) |
|---|---|
| State | Accepted, archived |
| Delegated to: | Rafael Wysocki |
| Headers | show |
diff --git a/drivers/acpi/ec.c b/drivers/acpi/ec.c index a94ee9f..3c97122 100644 --- a/drivers/acpi/ec.c +++ b/drivers/acpi/ec.c @@ -690,11 +690,21 @@ static int acpi_ec_sync_query(struct acpi_ec *ec, u8 *data) static void acpi_ec_gpe_query(void *ec_cxt) { struct acpi_ec *ec = ec_cxt; + acpi_status status; + u32 glk; if (!ec) return; mutex_lock(&ec->mutex); + if (ec->global_lock) { + status = acpi_acquire_global_lock(ACPI_EC_UDELAY_GLK, &glk); + if (ACPI_FAILURE(status)) + goto unlock; + } acpi_ec_sync_query(ec, NULL); + if (ec->global_lock) + acpi_release_global_lock(glk); +unlock: mutex_unlock(&ec->mutex); }
Currently QR_EC is queued up on CPU 0 to be safe with SMM because there is no global lock held for acpi_ec_gpe_query(). As we are about to move QR_EC to a non CPU 0 bound work queue to avoid invoking kmalloc() in advance_transaction(), we have to acquire global lock for the new QR_EC work item to avoid regressions. Known issue: 1. Global lock for acpi_ec_clear(). This is an existing issue that acpi_ec_clear() which invokes acpi_ec_sync_query() also suffers from the same issue. But this patch's target is only the code to invoke acpi_ec_sync_query() in a CPU 0 bound work queue item, and the acpi_ec_clear() can be automatically fixed by further patch that merges the redundant code, so it is left unchanged. Signed-off-by: Lv Zheng <lv.zheng@intel.com> --- drivers/acpi/ec.c | 10 ++++++++++ 1 file changed, 10 insertions(+)