Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / drivers / acpi / acpica / evgpeinit.c
diff --git a/kernel/drivers/acpi/acpica/evgpeinit.c b/kernel/drivers/acpi/acpica/evgpeinit.c
new file mode 100644 (file)
index 0000000..8840296
--- /dev/null
@@ -0,0 +1,446 @@
+/******************************************************************************
+ *
+ * Module Name: evgpeinit - System GPE initialization and update
+ *
+ *****************************************************************************/
+
+/*
+ * Copyright (C) 2000 - 2015, Intel Corp.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions, and the following disclaimer,
+ *    without modification.
+ * 2. Redistributions in binary form must reproduce at minimum a disclaimer
+ *    substantially similar to the "NO WARRANTY" disclaimer below
+ *    ("Disclaimer") and any redistribution must be conditioned upon
+ *    including a substantially similar Disclaimer requirement for further
+ *    binary redistribution.
+ * 3. Neither the names of the above-listed copyright holders nor the names
+ *    of any contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * Alternatively, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") version 2 as published by the Free
+ * Software Foundation.
+ *
+ * NO WARRANTY
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * HOLDERS OR CONTRIBUTORS BE LIABLE FOR SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+ * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
+ * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGES.
+ */
+
+#include <acpi/acpi.h>
+#include "accommon.h"
+#include "acevents.h"
+#include "acnamesp.h"
+
+#define _COMPONENT          ACPI_EVENTS
+ACPI_MODULE_NAME("evgpeinit")
+#if (!ACPI_REDUCED_HARDWARE)   /* Entire module */
+/*
+ * Note: History of _PRW support in ACPICA
+ *
+ * Originally (2000 - 2010), the GPE initialization code performed a walk of
+ * the entire namespace to execute the _PRW methods and detect all GPEs
+ * capable of waking the system.
+ *
+ * As of 10/2010, the _PRW method execution has been removed since it is
+ * actually unnecessary. The host OS must in fact execute all _PRW methods
+ * in order to identify the device/power-resource dependencies. We now put
+ * the onus on the host OS to identify the wake GPEs as part of this process
+ * and to inform ACPICA of these GPEs via the acpi_setup_gpe_for_wake interface. This
+ * not only reduces the complexity of the ACPICA initialization code, but in
+ * some cases (on systems with very large namespaces) it should reduce the
+ * kernel boot time as well.
+ */
+
+/*******************************************************************************
+ *
+ * FUNCTION:    acpi_ev_gpe_initialize
+ *
+ * PARAMETERS:  None
+ *
+ * RETURN:      Status
+ *
+ * DESCRIPTION: Initialize the GPE data structures and the FADT GPE 0/1 blocks
+ *
+ ******************************************************************************/
+acpi_status acpi_ev_gpe_initialize(void)
+{
+       u32 register_count0 = 0;
+       u32 register_count1 = 0;
+       u32 gpe_number_max = 0;
+       acpi_status status;
+
+       ACPI_FUNCTION_TRACE(ev_gpe_initialize);
+
+       ACPI_DEBUG_PRINT_RAW((ACPI_DB_INIT,
+                             "Initializing General Purpose Events (GPEs):\n"));
+
+       status = acpi_ut_acquire_mutex(ACPI_MTX_NAMESPACE);
+       if (ACPI_FAILURE(status)) {
+               return_ACPI_STATUS(status);
+       }
+
+       /*
+        * Initialize the GPE Block(s) defined in the FADT
+        *
+        * Why the GPE register block lengths are divided by 2:  From the ACPI
+        * Spec, section "General-Purpose Event Registers", we have:
+        *
+        * "Each register block contains two registers of equal length
+        *  GPEx_STS and GPEx_EN (where x is 0 or 1). The length of the
+        *  GPE0_STS and GPE0_EN registers is equal to half the GPE0_LEN
+        *  The length of the GPE1_STS and GPE1_EN registers is equal to
+        *  half the GPE1_LEN. If a generic register block is not supported
+        *  then its respective block pointer and block length values in the
+        *  FADT table contain zeros. The GPE0_LEN and GPE1_LEN do not need
+        *  to be the same size."
+        */
+
+       /*
+        * Determine the maximum GPE number for this machine.
+        *
+        * Note: both GPE0 and GPE1 are optional, and either can exist without
+        * the other.
+        *
+        * If EITHER the register length OR the block address are zero, then that
+        * particular block is not supported.
+        */
+       if (acpi_gbl_FADT.gpe0_block_length &&
+           acpi_gbl_FADT.xgpe0_block.address) {
+
+               /* GPE block 0 exists (has both length and address > 0) */
+
+               register_count0 = (u16)(acpi_gbl_FADT.gpe0_block_length / 2);
+               gpe_number_max =
+                   (register_count0 * ACPI_GPE_REGISTER_WIDTH) - 1;
+
+               /* Install GPE Block 0 */
+
+               status = acpi_ev_create_gpe_block(acpi_gbl_fadt_gpe_device,
+                                                 acpi_gbl_FADT.xgpe0_block.
+                                                 address,
+                                                 acpi_gbl_FADT.xgpe0_block.
+                                                 space_id, register_count0, 0,
+                                                 acpi_gbl_FADT.sci_interrupt,
+                                                 &acpi_gbl_gpe_fadt_blocks[0]);
+
+               if (ACPI_FAILURE(status)) {
+                       ACPI_EXCEPTION((AE_INFO, status,
+                                       "Could not create GPE Block 0"));
+               }
+       }
+
+       if (acpi_gbl_FADT.gpe1_block_length &&
+           acpi_gbl_FADT.xgpe1_block.address) {
+
+               /* GPE block 1 exists (has both length and address > 0) */
+
+               register_count1 = (u16)(acpi_gbl_FADT.gpe1_block_length / 2);
+
+               /* Check for GPE0/GPE1 overlap (if both banks exist) */
+
+               if ((register_count0) &&
+                   (gpe_number_max >= acpi_gbl_FADT.gpe1_base)) {
+                       ACPI_ERROR((AE_INFO,
+                                   "GPE0 block (GPE 0 to %u) overlaps the GPE1 block "
+                                   "(GPE %u to %u) - Ignoring GPE1",
+                                   gpe_number_max, acpi_gbl_FADT.gpe1_base,
+                                   acpi_gbl_FADT.gpe1_base +
+                                   ((register_count1 *
+                                     ACPI_GPE_REGISTER_WIDTH) - 1)));
+
+                       /* Ignore GPE1 block by setting the register count to zero */
+
+                       register_count1 = 0;
+               } else {
+                       /* Install GPE Block 1 */
+
+                       status =
+                           acpi_ev_create_gpe_block(acpi_gbl_fadt_gpe_device,
+                                                    acpi_gbl_FADT.xgpe1_block.
+                                                    address,
+                                                    acpi_gbl_FADT.xgpe1_block.
+                                                    space_id, register_count1,
+                                                    acpi_gbl_FADT.gpe1_base,
+                                                    acpi_gbl_FADT.
+                                                    sci_interrupt,
+                                                    &acpi_gbl_gpe_fadt_blocks
+                                                    [1]);
+
+                       if (ACPI_FAILURE(status)) {
+                               ACPI_EXCEPTION((AE_INFO, status,
+                                               "Could not create GPE Block 1"));
+                       }
+
+                       /*
+                        * GPE0 and GPE1 do not have to be contiguous in the GPE number
+                        * space. However, GPE0 always starts at GPE number zero.
+                        */
+                       gpe_number_max = acpi_gbl_FADT.gpe1_base +
+                           ((register_count1 * ACPI_GPE_REGISTER_WIDTH) - 1);
+               }
+       }
+
+       /* Exit if there are no GPE registers */
+
+       if ((register_count0 + register_count1) == 0) {
+
+               /* GPEs are not required by ACPI, this is OK */
+
+               ACPI_DEBUG_PRINT((ACPI_DB_INIT,
+                                 "There are no GPE blocks defined in the FADT\n"));
+               status = AE_OK;
+               goto cleanup;
+       }
+
+cleanup:
+       (void)acpi_ut_release_mutex(ACPI_MTX_NAMESPACE);
+       return_ACPI_STATUS(AE_OK);
+}
+
+/*******************************************************************************
+ *
+ * FUNCTION:    acpi_ev_update_gpes
+ *
+ * PARAMETERS:  table_owner_id      - ID of the newly-loaded ACPI table
+ *
+ * RETURN:      None
+ *
+ * DESCRIPTION: Check for new GPE methods (_Lxx/_Exx) made available as a
+ *              result of a Load() or load_table() operation. If new GPE
+ *              methods have been installed, register the new methods.
+ *
+ ******************************************************************************/
+
+void acpi_ev_update_gpes(acpi_owner_id table_owner_id)
+{
+       struct acpi_gpe_xrupt_info *gpe_xrupt_info;
+       struct acpi_gpe_block_info *gpe_block;
+       struct acpi_gpe_walk_info walk_info;
+       acpi_status status = AE_OK;
+
+       /*
+        * Find any _Lxx/_Exx GPE methods that have just been loaded.
+        *
+        * Any GPEs that correspond to new _Lxx/_Exx methods are immediately
+        * enabled.
+        *
+        * Examine the namespace underneath each gpe_device within the
+        * gpe_block lists.
+        */
+       status = acpi_ut_acquire_mutex(ACPI_MTX_EVENTS);
+       if (ACPI_FAILURE(status)) {
+               return;
+       }
+
+       walk_info.count = 0;
+       walk_info.owner_id = table_owner_id;
+       walk_info.execute_by_owner_id = TRUE;
+
+       /* Walk the interrupt level descriptor list */
+
+       gpe_xrupt_info = acpi_gbl_gpe_xrupt_list_head;
+       while (gpe_xrupt_info) {
+
+               /* Walk all Gpe Blocks attached to this interrupt level */
+
+               gpe_block = gpe_xrupt_info->gpe_block_list_head;
+               while (gpe_block) {
+                       walk_info.gpe_block = gpe_block;
+                       walk_info.gpe_device = gpe_block->node;
+
+                       status = acpi_ns_walk_namespace(ACPI_TYPE_METHOD,
+                                                       walk_info.gpe_device,
+                                                       ACPI_UINT32_MAX,
+                                                       ACPI_NS_WALK_NO_UNLOCK,
+                                                       acpi_ev_match_gpe_method,
+                                                       NULL, &walk_info, NULL);
+                       if (ACPI_FAILURE(status)) {
+                               ACPI_EXCEPTION((AE_INFO, status,
+                                               "While decoding _Lxx/_Exx methods"));
+                       }
+
+                       gpe_block = gpe_block->next;
+               }
+
+               gpe_xrupt_info = gpe_xrupt_info->next;
+       }
+
+       if (walk_info.count) {
+               ACPI_INFO((AE_INFO, "Enabled %u new GPEs", walk_info.count));
+       }
+
+       (void)acpi_ut_release_mutex(ACPI_MTX_EVENTS);
+       return;
+}
+
+/*******************************************************************************
+ *
+ * FUNCTION:    acpi_ev_match_gpe_method
+ *
+ * PARAMETERS:  Callback from walk_namespace
+ *
+ * RETURN:      Status
+ *
+ * DESCRIPTION: Called from acpi_walk_namespace. Expects each object to be a
+ *              control method under the _GPE portion of the namespace.
+ *              Extract the name and GPE type from the object, saving this
+ *              information for quick lookup during GPE dispatch. Allows a
+ *              per-owner_id evaluation if execute_by_owner_id is TRUE in the
+ *              walk_info parameter block.
+ *
+ *              The name of each GPE control method is of the form:
+ *              "_Lxx" or "_Exx", where:
+ *                  L      - means that the GPE is level triggered
+ *                  E      - means that the GPE is edge triggered
+ *                  xx     - is the GPE number [in HEX]
+ *
+ * If walk_info->execute_by_owner_id is TRUE, we only execute examine GPE methods
+ * with that owner.
+ *
+ ******************************************************************************/
+
+acpi_status
+acpi_ev_match_gpe_method(acpi_handle obj_handle,
+                        u32 level, void *context, void **return_value)
+{
+       struct acpi_namespace_node *method_node =
+           ACPI_CAST_PTR(struct acpi_namespace_node, obj_handle);
+       struct acpi_gpe_walk_info *walk_info =
+           ACPI_CAST_PTR(struct acpi_gpe_walk_info, context);
+       struct acpi_gpe_event_info *gpe_event_info;
+       u32 gpe_number;
+       char name[ACPI_NAME_SIZE + 1];
+       u8 type;
+
+       ACPI_FUNCTION_TRACE(ev_match_gpe_method);
+
+       /* Check if requested owner_id matches this owner_id */
+
+       if ((walk_info->execute_by_owner_id) &&
+           (method_node->owner_id != walk_info->owner_id)) {
+               return_ACPI_STATUS(AE_OK);
+       }
+
+       /*
+        * Match and decode the _Lxx and _Exx GPE method names
+        *
+        * 1) Extract the method name and null terminate it
+        */
+       ACPI_MOVE_32_TO_32(name, &method_node->name.integer);
+       name[ACPI_NAME_SIZE] = 0;
+
+       /* 2) Name must begin with an underscore */
+
+       if (name[0] != '_') {
+               return_ACPI_STATUS(AE_OK);      /* Ignore this method */
+       }
+
+       /*
+        * 3) Edge/Level determination is based on the 2nd character
+        *    of the method name
+        */
+       switch (name[1]) {
+       case 'L':
+
+               type = ACPI_GPE_LEVEL_TRIGGERED;
+               break;
+
+       case 'E':
+
+               type = ACPI_GPE_EDGE_TRIGGERED;
+               break;
+
+       default:
+
+               /* Unknown method type, just ignore it */
+
+               ACPI_DEBUG_PRINT((ACPI_DB_LOAD,
+                                 "Ignoring unknown GPE method type: %s "
+                                 "(name not of form _Lxx or _Exx)", name));
+               return_ACPI_STATUS(AE_OK);
+       }
+
+       /* 4) The last two characters of the name are the hex GPE Number */
+
+       gpe_number = ACPI_STRTOUL(&name[2], NULL, 16);
+       if (gpe_number == ACPI_UINT32_MAX) {
+
+               /* Conversion failed; invalid method, just ignore it */
+
+               ACPI_DEBUG_PRINT((ACPI_DB_LOAD,
+                                 "Could not extract GPE number from name: %s "
+                                 "(name is not of form _Lxx or _Exx)", name));
+               return_ACPI_STATUS(AE_OK);
+       }
+
+       /* Ensure that we have a valid GPE number for this GPE block */
+
+       gpe_event_info =
+           acpi_ev_low_get_gpe_info(gpe_number, walk_info->gpe_block);
+       if (!gpe_event_info) {
+               /*
+                * This gpe_number is not valid for this GPE block, just ignore it.
+                * However, it may be valid for a different GPE block, since GPE0
+                * and GPE1 methods both appear under \_GPE.
+                */
+               return_ACPI_STATUS(AE_OK);
+       }
+
+       if ((ACPI_GPE_DISPATCH_TYPE(gpe_event_info->flags) ==
+            ACPI_GPE_DISPATCH_HANDLER) ||
+           (ACPI_GPE_DISPATCH_TYPE(gpe_event_info->flags) ==
+            ACPI_GPE_DISPATCH_RAW_HANDLER)) {
+
+               /* If there is already a handler, ignore this GPE method */
+
+               return_ACPI_STATUS(AE_OK);
+       }
+
+       if (ACPI_GPE_DISPATCH_TYPE(gpe_event_info->flags) ==
+           ACPI_GPE_DISPATCH_METHOD) {
+               /*
+                * If there is already a method, ignore this method. But check
+                * for a type mismatch (if both the _Lxx AND _Exx exist)
+                */
+               if (type != (gpe_event_info->flags & ACPI_GPE_XRUPT_TYPE_MASK)) {
+                       ACPI_ERROR((AE_INFO,
+                                   "For GPE 0x%.2X, found both _L%2.2X and _E%2.2X methods",
+                                   gpe_number, gpe_number, gpe_number));
+               }
+               return_ACPI_STATUS(AE_OK);
+       }
+
+       /* Disable the GPE in case it's been enabled already. */
+
+       (void)acpi_hw_low_set_gpe(gpe_event_info, ACPI_GPE_DISABLE);
+
+       /*
+        * Add the GPE information from above to the gpe_event_info block for
+        * use during dispatch of this GPE.
+        */
+       gpe_event_info->flags &= ~(ACPI_GPE_DISPATCH_MASK);
+       gpe_event_info->flags |= (u8)(type | ACPI_GPE_DISPATCH_METHOD);
+       gpe_event_info->dispatch.method_node = method_node;
+
+       ACPI_DEBUG_PRINT((ACPI_DB_LOAD,
+                         "Registered GPE method %s as GPE number 0x%.2X\n",
+                         name, gpe_number));
+       return_ACPI_STATUS(AE_OK);
+}
+
+#endif                         /* !ACPI_REDUCED_HARDWARE */