arm-trusted-firmware/lib/cpus/aarch64/neoverse_v3.S
Boyan Karatotev 89dba82dfa perf(cpus): make reset errata do fewer branches
Errata application is painful for performance. For a start, it's done
when the core has just come out of reset, which means branch predictors
and caches will be empty so a branch to a workaround function must be
fetched from memory and that round trip is very slow. Then it also runs
with the I-cache off, which means that the loop to iterate over the
workarounds must also be fetched from memory on each iteration.

We can remove both branches. First, we can simply apply every erratum
directly instead of defining a workaround function and jumping to it.
Currently, no errata that need to be applied at both reset and runtime,
with the same workaround function, exist. If the need arose in future,
this should be achievable with a reset + runtime wrapper combo.

Then, we can construct a function that applies each erratum linearly
instead of looping over the list. If this function is part of the reset
function, then the only "far" branches at reset will be for the checker
functions. Importantly, this mitigates the slowdown even when an erratum
is disabled.

The result is ~50% speedup on N1SDP and ~20% on AArch64 Juno on wakeup
from PSCI calls that end in powerdown. This is roughly back to the
baseline of v2.9, before the errata framework regressed on performance
(or a little better). It is important to note that there are other
slowdowns since then that remain unknown.

Change-Id: Ie4d5288a331b11fd648e5c4a0b652b74160b07b9
Signed-off-by: Boyan Karatotev <boyan.karatotev@arm.com>
2025-02-24 09:36:11 +00:00

133 lines
4.1 KiB
ArmAsm

/*
* Copyright (c) 2022-2025, Arm Limited. All rights reserved.
*
* SPDX-License-Identifier: BSD-3-Clause
*/
#include <arch.h>
#include <asm_macros.S>
#include <common/bl_common.h>
#include <neoverse_v3.h>
#include <cpu_macros.S>
#include <plat_macros.S>
#include "wa_cve_2022_23960_bhb_vector.S"
/* Hardware handled coherency */
#if HW_ASSISTED_COHERENCY == 0
#error "Neoverse V3 must be compiled with HW_ASSISTED_COHERENCY enabled"
#endif
/* 64-bit only core */
#if CTX_INCLUDE_AARCH32_REGS == 1
#error "Neoverse V3 supports only AArch64. Compile with CTX_INCLUDE_AARCH32_REGS=0"
#endif
cpu_reset_prologue neoverse_v3
.global check_erratum_neoverse_v3_3701767
add_erratum_entry neoverse_v3, ERRATUM(3701767), ERRATA_V3_3701767
check_erratum_ls neoverse_v3, ERRATUM(3701767), CPU_REV(0, 2)
workaround_reset_start neoverse_v3, ERRATUM(2970647), ERRATA_V3_2970647
/* Add ISB before MRS reads of MPIDR_EL1/MIDR_EL1 */
ldr x0, =0x1
msr S3_6_c15_c8_0, x0 /* msr CPUPSELR_EL3, X0 */
ldr x0, =0xd5380000
msr S3_6_c15_c8_2, x0 /* msr CPUPOR_EL3, X0 */
ldr x0, =0xFFFFFF40
msr S3_6_c15_c8_3,x0 /* msr CPUPMR_EL3, X0 */
ldr x0, =0x000080010033f
msr S3_6_c15_c8_1, x0 /* msr CPUPCR_EL3, X0 */
isb
workaround_reset_end neoverse_v3, ERRATUM(2970647)
check_erratum_ls neoverse_v3, ERRATUM(2970647), CPU_REV(0, 0)
#if WORKAROUND_CVE_2022_23960
wa_cve_2022_23960_bhb_vector_table NEOVERSE_V3_BHB_LOOP_COUNT, neoverse_v3
#endif /* WORKAROUND_CVE_2022_23960 */
/* Disable hardware page aggregation. Enables mitigation for `CVE-2024-5660` */
workaround_reset_start neoverse_v3, CVE(2024, 5660), WORKAROUND_CVE_2024_5660
sysreg_bit_set NEOVERSE_V3_CPUECTLR_EL1, BIT(46)
workaround_reset_end neoverse_v3, CVE(2024, 5660)
check_erratum_ls neoverse_v3, CVE(2024, 5660), CPU_REV(0, 1)
workaround_reset_start neoverse_v3, CVE(2022,23960), WORKAROUND_CVE_2022_23960
#if IMAGE_BL31
/*
* The Neoverse V3 generic vectors are overridden to apply errata
* mitigation on exception entry from lower ELs.
*/
override_vector_table wa_cve_vbar_neoverse_v3
#endif /* IMAGE_BL31 */
workaround_reset_end neoverse_v3, CVE(2022,23960)
check_erratum_chosen neoverse_v3, CVE(2022, 23960), WORKAROUND_CVE_2022_23960
workaround_reset_start neoverse_v3, CVE(2024, 7881), WORKAROUND_CVE_2024_7881
/* ---------------------------------
* Sets BIT41 of CPUACTLR6_EL1 which
* disables L1 Data cache prefetcher
* ---------------------------------
*/
sysreg_bit_set NEOVERSE_V3_CPUACTLR6_EL1, BIT(41)
workaround_reset_end neoverse_v3, CVE(2024, 7881)
check_erratum_chosen neoverse_v3, CVE(2024, 7881), WORKAROUND_CVE_2024_7881
/* ---------------------------------------------
* HW will do the cache maintenance while powering down
* ---------------------------------------------
*/
func neoverse_v3_core_pwr_dwn
/* ---------------------------------------------
* Enable CPU power down bit in power control register
* ---------------------------------------------
*/
sysreg_bit_set NEOVERSE_V3_CPUPWRCTLR_EL1, \
NEOVERSE_V3_CPUPWRCTLR_EL1_CORE_PWRDN_BIT
isb
ret
endfunc neoverse_v3_core_pwr_dwn
cpu_reset_func_start neoverse_v3
/* Disable speculative loads */
msr SSBS, xzr
cpu_reset_func_end neoverse_v3
/* ---------------------------------------------
* This function provides Neoverse V3 specific
* register information for crash reporting.
* It needs to return with x6 pointing to
* a list of register names in ascii and
* x8 - x15 having values of registers to be
* reported.
* ---------------------------------------------
*/
.section .rodata.neoverse_v3_regs, "aS"
neoverse_v3_regs: /* The ascii list of register names to be reported */
.asciz "cpuectlr_el1", ""
func neoverse_v3_cpu_reg_dump
adr x6, neoverse_v3_regs
mrs x8, NEOVERSE_V3_CPUECTLR_EL1
ret
endfunc neoverse_v3_cpu_reg_dump
declare_cpu_ops neoverse_v3, NEOVERSE_V3_VNAE_MIDR, \
neoverse_v3_reset_func, \
neoverse_v3_core_pwr_dwn
declare_cpu_ops_wa_4 neoverse_v3, NEOVERSE_V3_MIDR, \
neoverse_v3_reset_func, \
CPU_NO_EXTRA1_FUNC, \
CPU_NO_EXTRA2_FUNC, \
CPU_NO_EXTRA3_FUNC, \
check_erratum_neoverse_v3_7881, \
neoverse_v3_core_pwr_dwn