Move ID_AA64SMFR0_EL1 definitions before ID_AA64ZFR0_EL1, and fix formatting in identcpu.c
- Queries
- All Stories
- Search
- Advanced Search
- Transactions
- Transaction Logs
Advanced Search
Thu, Mar 13
Wed, Mar 5
Fri, Feb 28
Mon, Feb 24
Feb 13 2025
Small changes as a result of comments on parent change
Add definition for ETS 0b1 value, renamed S*ERR and A*ERR values, and re-ordered MTE_frac definitions.
Feb 4 2025
Simplify arm_tmr_do_delay() further.
Feb 3 2025
The timer register is 64-bit, so best to use *int64_t, rather than *int32_t.
Jan 29 2025
Amended arm_tmr_do_delay() to use end for wfet(), used get_kernel_reg() rather than READ_SPECIALREG(), and added CPU_FEAT_AFTER_DEV flag, so support is checked after secondary cpus are brought up.
Use the CPU feature framework to determine whether the WFxT instructions are available on all CPUs, and set enable_wfxt accordingly. When the timer is used for DELAY(), it'll check whether it can use wfet().
Jan 24 2025
Jan 22 2025
Move configuration of hcrx_el2 to avoid trashing cptr_el2
Addressed review comments
Moved HCRX register definition to hypervisor.h, moved initial
configuration of hcrx_el2 to after .Ldone_vhe
Jan 21 2025
In D48584#1108156, @andrew wrote:Does the value of the register matter when no guest is running? Does the hardware change the register when a guest is running? If not we can restrict when we save & restore it, e.g. if the host doesn't care what value it has, and it's never changed by hardware then we only need to restore it for guests and can drop the vmm_hyp_reg_store change.
Cast to uint64_t not needed, removed.