/* * Copyright 2020 Advanced Micro Devices, Inc. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), * to deal in the Software without restriction, including without limitation * the rights to use, copy, modify, merge, publish, distribute, sublicense, * and/or sell copies of the Software, and to permit persons to whom the * Software is furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice shall be included in * all copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR * OTHER DEALINGS IN THE SOFTWARE.
*/
/* * DO NOT use these for err/warn/info/debug messages. * Use dev_err, dev_warn, dev_info and dev_dbg instead. * They are more MGPU friendly.
*/ #undef pr_err #undef pr_warn #undef pr_info #undef pr_debug
/* Redefine the SMU error codes here. * * Note that these definitions are redundant and should be removed * when the SMU has exported a unified header file containing these * macros, which header file we can just include and use the SMU's * macros. At the moment, these error codes are defined by the SMU * per-ASIC unfortunately, yet we're a one driver for all ASICs.
*/ #define SMU_RESP_NONE 0 #define SMU_RESP_OK 1 #define SMU_RESP_CMD_FAIL 0xFF #define SMU_RESP_CMD_UNKNOWN 0xFE #define SMU_RESP_CMD_BAD_PREREQ 0xFD #define SMU_RESP_BUSY_OTHER 0xFC #define SMU_RESP_DEBUG_END 0xFB
#define SMU_RESP_UNEXP (~0U) /** * __smu_cmn_poll_stat -- poll for a status from the SMU * @smu: a pointer to SMU context * * Returns the status of the SMU, which could be, * 0, the SMU is busy with your command; * 1, execution status: success, execution result: success; * 0xFF, execution status: success, execution result: failure; * 0xFE, unknown command; * 0xFD, valid command, but bad (command) prerequisites; * 0xFC, the command was rejected as the SMU is busy; * 0xFB, "SMC_Result_DebugDataDumpEnd". * * The values here are not defined by macros, because I'd rather we * include a single header file which defines them, which is * maintained by the SMU FW team, so that we're impervious to firmware * changes. At the moment those values are defined in various header * files, one for each ASIC, yet here we're a single ASIC-agnostic * interface. Such a change can be followed-up by a subsequent patch.
*/ static u32 __smu_cmn_poll_stat(struct smu_context *smu)
{ struct amdgpu_device *adev = smu->adev; int timeout = adev->usec_timeout * 20;
u32 reg;
for ( ; timeout > 0; timeout--) {
reg = RREG32(smu->resp_reg); if ((reg & MP1_C2PMSG_90__CONTENT_MASK) != 0) break;
switch (reg_c2pmsg_90) { case SMU_RESP_NONE: {
msg_idx = RREG32(smu->msg_reg);
prm = RREG32(smu->param_reg);
dev_err_ratelimited(adev->dev, "SMU: I'm not done with your previous command: SMN_C2PMSG_66:0x%08X SMN_C2PMSG_82:0x%08X",
msg_idx, prm);
} break; case SMU_RESP_OK: /* The SMU executed the command. It completed with a * successful result.
*/ break; case SMU_RESP_CMD_FAIL: /* The SMU executed the command. It completed with an * unsuccessful result.
*/ break; case SMU_RESP_CMD_UNKNOWN:
dev_err_ratelimited(adev->dev, "SMU: unknown command: index:%d param:0x%08X message:%s",
msg_index, param, message); break; case SMU_RESP_CMD_BAD_PREREQ:
dev_err_ratelimited(adev->dev, "SMU: valid command, bad prerequisites: index:%d param:0x%08X message:%s",
msg_index, param, message); break; case SMU_RESP_BUSY_OTHER:
dev_err_ratelimited(adev->dev, "SMU: I'm very busy for your command: index:%d param:0x%08X message:%s",
msg_index, param, message); break; case SMU_RESP_DEBUG_END:
dev_err_ratelimited(adev->dev, "SMU: I'm debugging!"); break; case SMU_RESP_UNEXP: if (amdgpu_device_bus_status_check(smu->adev)) { /* print error immediately if device is off the bus */
dev_err(adev->dev, "SMU: response:0x%08X for index:%d param:0x%08X message:%s?",
reg_c2pmsg_90, msg_index, param, message); break;
}
fallthrough; default:
dev_err_ratelimited(adev->dev, "SMU: response:0x%08X for index:%d param:0x%08X message:%s?",
reg_c2pmsg_90, msg_index, param, message); break;
}
}
staticint __smu_cmn_reg2errno(struct smu_context *smu, u32 reg_c2pmsg_90)
{ int res;
switch (reg_c2pmsg_90) { case SMU_RESP_NONE: /* The SMU is busy--still executing your command.
*/
res = -ETIME; break; case SMU_RESP_OK:
res = 0; break; case SMU_RESP_CMD_FAIL: /* Command completed successfully, but the command * status was failure.
*/
res = -EIO; break; case SMU_RESP_CMD_UNKNOWN: /* Unknown command--ignored by the SMU.
*/
res = -EOPNOTSUPP; break; case SMU_RESP_CMD_BAD_PREREQ: /* Valid command--bad prerequisites.
*/
res = -EINVAL; break; case SMU_RESP_BUSY_OTHER: /* The SMU is busy with other commands. The client * should retry in 10 us.
*/
res = -EBUSY; break; default: /* Unknown or debug response from the SMU.
*/
res = -EREMOTEIO; break;
}
/* When there is RAS fatal error, FW won't process non-RAS priority * messages. Don't allow any messages other than RAS priority messages.
*/
fed_status = amdgpu_ras_get_fed_status(adev); if (fed_status) { if (!(flags & SMU_MSG_RAS_PRI)) {
dev_dbg(adev->dev, "RAS error detected, skip sending %s",
smu_get_message_name(smu, msg)); return -EACCES;
}
/* FW will ignore non-priority messages when a RAS fatal error * is detected. Hence it is possible that a previous message * wouldn't have got response. Allow to continue without polling * for response status for priority messages.
*/
resp = RREG32(smu->resp_reg);
dev_dbg(adev->dev, "Sending RAS priority message %s response status: %x",
smu_get_message_name(smu, msg), resp); if (resp == 0)
*poll = false;
}
return 0;
} /** * smu_cmn_send_msg_without_waiting -- send the message; don't wait for status * @smu: pointer to an SMU context * @msg_index: message index * @param: message parameter to send to the SMU * * Send a message to the SMU with the parameter passed. Do not wait * for status/result of the message, thus the "without_waiting". * * Return 0 on success, -errno on error if we weren't able to _send_ * the message for some reason. See __smu_cmn_reg2errno() for details * of the -errno.
*/ int smu_cmn_send_msg_without_waiting(struct smu_context *smu,
uint16_t msg_index,
uint32_t param)
{ struct amdgpu_device *adev = smu->adev;
u32 reg; int res;
if (adev->no_hw_access) return 0;
if (smu->smc_fw_state == SMU_FW_HANG) {
dev_err(adev->dev, "SMU is in hanged state, failed to send smu message!\n");
res = -EREMOTEIO; goto Out;
}
if (smu->smc_fw_state == SMU_FW_INIT) {
smu->smc_fw_state = SMU_FW_RUNTIME;
} else {
reg = __smu_cmn_poll_stat(smu);
res = __smu_cmn_reg2errno(smu, reg); if (reg == SMU_RESP_NONE || res == -EREMOTEIO) goto Out;
}
__smu_cmn_send_msg(smu, msg_index, param);
res = 0;
Out: if (unlikely(adev->pm.smu_debug_mask & SMU_DEBUG_HALT_ON_ERROR) &&
res && (res != -ETIME)) {
amdgpu_device_halt(adev);
WARN_ON(1);
}
return res;
}
/** * smu_cmn_wait_for_response -- wait for response from the SMU * @smu: pointer to an SMU context * * Wait for status from the SMU. * * Return 0 on success, -errno on error, indicating the execution * status and result of the message being waited for. See * __smu_cmn_reg2errno() for details of the -errno.
*/ int smu_cmn_wait_for_response(struct smu_context *smu)
{
u32 reg; int res;
reg = __smu_cmn_poll_stat(smu);
res = __smu_cmn_reg2errno(smu, reg);
if (res == -EREMOTEIO)
smu->smc_fw_state = SMU_FW_HANG;
if (unlikely(smu->adev->pm.smu_debug_mask & SMU_DEBUG_HALT_ON_ERROR) &&
res && (res != -ETIME)) {
amdgpu_device_halt(smu->adev);
WARN_ON(1);
}
return res;
}
/** * smu_cmn_send_smc_msg_with_param -- send a message with parameter * @smu: pointer to an SMU context * @msg: message to send * @param: parameter to send to the SMU * @read_arg: pointer to u32 to return a value from the SMU back * to the caller * * Send the message @msg with parameter @param to the SMU, wait for * completion of the command, and return back a value from the SMU in * @read_arg pointer. * * Return 0 on success, -errno when a problem is encountered sending * message or receiving reply. If there is a PCI bus recovery or * the destination is a virtual GPU which does not allow this message * type, the message is simply dropped and success is also returned. * See __smu_cmn_reg2errno() for details of the -errno. * * If we weren't able to send the message to the SMU, we also print * the error to the standard log. * * Command completion status is printed only if the -errno is * -EREMOTEIO, indicating that the SMU returned back an * undefined/unknown/unspecified result. All other cases are * well-defined, not printed, but instead given back to the client to * decide what further to do. * * The return value, @read_arg is read back regardless, to give back * more information to the client, which on error would most likely be * @param, but we can't assume that. This also eliminates more * conditionals.
*/ int smu_cmn_send_smc_msg_with_param(struct smu_context *smu, enum smu_message_type msg,
uint32_t param,
uint32_t *read_arg)
{ struct amdgpu_device *adev = smu->adev; int res, index; bool poll = true;
u32 reg;
if (adev->no_hw_access) return 0;
index = smu_cmn_to_asic_specific_index(smu,
CMN2ASIC_MAPPING_MSG,
msg); if (index < 0) return index == -EACCES ? 0 : index;
mutex_lock(&smu->message_lock);
if (smu->smc_fw_caps & SMU_FW_CAP_RAS_PRI) {
res = __smu_cmn_ras_filter_msg(smu, msg, &poll); if (res) goto Out;
}
if (smu->smc_fw_state == SMU_FW_HANG) {
dev_err(adev->dev, "SMU is in hanged state, failed to send smu message!\n");
res = -EREMOTEIO; goto Out;
} elseif (smu->smc_fw_state == SMU_FW_INIT) { /* Ignore initial smu response register value */
poll = false;
smu->smc_fw_state = SMU_FW_RUNTIME;
}
int smu_cmn_feature_is_enabled(struct smu_context *smu, enum smu_feature_mask mask)
{ struct amdgpu_device *adev = smu->adev;
uint64_t enabled_features; int feature_id;
if (__smu_get_enabled_features(smu, &enabled_features)) {
dev_err(adev->dev, "Failed to retrieve enabled ppfeatures!\n"); return 0;
}
/* * For Renoir and Cyan Skillfish, they are assumed to have all features * enabled. Also considering they have no feature_map available, the * check here can avoid unwanted feature_map check below.
*/ if (enabled_features == ULLONG_MAX) return 1;
switch (clk_type) { case SMU_MCLK: case SMU_UCLK:
feature_id = SMU_FEATURE_DPM_UCLK_BIT; break; case SMU_GFXCLK: case SMU_SCLK:
feature_id = SMU_FEATURE_DPM_GFXCLK_BIT; break; case SMU_SOCCLK:
feature_id = SMU_FEATURE_DPM_SOCCLK_BIT; break; case SMU_VCLK: case SMU_VCLK1:
feature_id = SMU_FEATURE_DPM_VCLK_BIT; break; case SMU_DCLK: case SMU_DCLK1:
feature_id = SMU_FEATURE_DPM_DCLK_BIT; break; case SMU_FCLK:
feature_id = SMU_FEATURE_DPM_FCLK_BIT; break; default: returntrue;
}
if (!smu_cmn_feature_is_enabled(smu, feature_id)) returnfalse;
returntrue;
}
int smu_cmn_get_enabled_mask(struct smu_context *smu,
uint64_t *feature_mask)
{
uint32_t *feature_mask_high;
uint32_t *feature_mask_low; int ret = 0, index = 0;
index = smu_cmn_to_asic_specific_index(smu,
CMN2ASIC_MAPPING_MSG,
SMU_MSG_GetEnabledSmuFeatures); if (index > 0) {
ret = smu_cmn_send_smc_msg_with_param(smu,
SMU_MSG_GetEnabledSmuFeatures,
0,
feature_mask_low); if (ret) return ret;
ret = smu_cmn_send_smc_msg_with_param(smu,
SMU_MSG_GetEnabledSmuFeatures,
1,
feature_mask_high);
} else {
ret = smu_cmn_send_smc_msg(smu,
SMU_MSG_GetEnabledSmuFeaturesHigh,
feature_mask_high); if (ret) return ret;
ret = smu_cmn_send_smc_msg(smu,
SMU_MSG_GetEnabledSmuFeaturesLow,
feature_mask_low);
}
for (i = 0; i < SMU_FEATURE_COUNT; i++) {
feature_index = smu_cmn_to_asic_specific_index(smu,
CMN2ASIC_MAPPING_FEATURE,
i); if (feature_index < 0) continue;
if (feature_2_enabled) {
ret = smu_cmn_feature_update_enable_state(smu,
feature_2_enabled, true); if (ret) return ret;
} if (feature_2_disabled) {
ret = smu_cmn_feature_update_enable_state(smu,
feature_2_disabled, false); if (ret) return ret;
}
return ret;
}
/** * smu_cmn_disable_all_features_with_exception - disable all dpm features * except this specified by * @mask * * @smu: smu_context pointer * @mask: the dpm feature which should not be disabled * SMU_FEATURE_COUNT: no exception, all dpm features * to disable * * Returns: * 0 on success or a negative error code on failure.
*/ int smu_cmn_disable_all_features_with_exception(struct smu_context *smu, enum smu_feature_mask mask)
{
uint64_t features_to_disable = U64_MAX; int skipped_feature_id;
if (mask != SMU_FEATURE_COUNT) {
skipped_feature_id = smu_cmn_to_asic_specific_index(smu,
CMN2ASIC_MAPPING_FEATURE,
mask); if (skipped_feature_id < 0) return -EINVAL;
int smu_cmn_get_smc_version(struct smu_context *smu,
uint32_t *if_version,
uint32_t *smu_version)
{ int ret = 0;
if (!if_version && !smu_version) return -EINVAL;
if (smu->smc_fw_if_version && smu->smc_fw_version)
{ if (if_version)
*if_version = smu->smc_fw_if_version;
if (smu_version)
*smu_version = smu->smc_fw_version;
return 0;
}
if (if_version) {
ret = smu_cmn_send_smc_msg(smu, SMU_MSG_GetDriverIfVersion, if_version); if (ret) return ret;
smu->smc_fw_if_version = *if_version;
}
if (smu_version) {
ret = smu_cmn_send_smc_msg(smu, SMU_MSG_GetSmuVersion, smu_version); if (ret) return ret;
smu->smc_fw_version = *smu_version;
}
return ret;
}
int smu_cmn_update_table(struct smu_context *smu, enum smu_table_id table_index, int argument, void *table_data, bool drv2smu)
{ struct smu_table_context *smu_table = &smu->smu_table; struct amdgpu_device *adev = smu->adev; struct smu_table *table = &smu_table->driver_table; int table_id = smu_cmn_to_asic_specific_index(smu,
CMN2ASIC_MAPPING_TABLE,
table_index);
uint32_t table_size; int ret = 0; if (!table_data || table_index >= SMU_TABLE_COUNT || table_id < 0) return -EINVAL;
table_size = smu_table->tables[table_index].size;
if (drv2smu) {
memcpy(table->cpu_addr, table_data, table_size); /* * Flush hdp cache: to guard the content seen by * GPU is consitent with CPU.
*/
amdgpu_asic_flush_hdp(adev, NULL);
}
ret = smu_cmn_send_smc_msg_with_param(smu, drv2smu ?
SMU_MSG_TransferTableDram2Smu :
SMU_MSG_TransferTableSmu2Dram,
table_id | ((argument & 0xFFFF) << 16),
NULL); if (ret) return ret;
if (!drv2smu) {
amdgpu_asic_invalidate_hdp(adev, NULL);
memcpy(table_data, table->cpu_addr, table_size);
}
return 0;
}
int smu_cmn_write_watermarks_table(struct smu_context *smu)
{ void *watermarks_table = smu->smu_table.watermarks_table;
/* * If the ASIC comes with no audio function, we always assume * it is "enabled".
*/
p = pci_get_domain_bus_and_slot(pci_domain_nr(adev->pdev->bus),
adev->pdev->bus->number, 1); if (!p) returntrue;
staticchar *smu_soc_policy_get_desc(struct smu_dpm_policy *policy, int level)
{ if (level < 0 || !(policy->level_mask & BIT(level))) return"Invalid";
switch (level) { case SOC_PSTATE_DEFAULT: return"soc_pstate_default"; case SOC_PSTATE_0: return"soc_pstate_0"; case SOC_PSTATE_1: return"soc_pstate_1"; case SOC_PSTATE_2: return"soc_pstate_2";
}
staticchar *smu_xgmi_plpd_policy_get_desc(struct smu_dpm_policy *policy, int level)
{ if (level < 0 || !(policy->level_mask & BIT(level))) return"Invalid";
switch (level) { case XGMI_PLPD_DISALLOW: return"plpd_disallow"; case XGMI_PLPD_DEFAULT: return"plpd_default"; case XGMI_PLPD_OPTIMIZED: return"plpd_optimized";
}
Die Informationen auf dieser Webseite wurden
nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit,
noch Qualität der bereit gestellten Informationen zugesichert.
Bemerkung:
Die farbliche Syntaxdarstellung und die Messung sind noch experimentell.