Time to drop wrmsrl()/rdmsrl()?

Time to drop wrmsrl()/rdmsrl()?

[Xin Li]

Hi Ingo,

Keeping them around only causes churn now, so probably it is time to drop
them?

https://lore.kernel.org/lkml/aK8fmMGH0rB2LuA9@google.com/

https://lore.kernel.org/lkml/aR0vK3z1owwM8X8H@intel.com/


I think all of their uses were changed to wrmsrq()/rdmsrq(), but they are
coming back, and with the latest Linux 6.19-rc1 tag, I get:

$git grep -w rdmsrl
arch/x86/events/amd/uncore.c:   rdmsrl(hwc->event_base, new);
arch/x86/include/asm/msr.h:#define rdmsrl(msr, val) rdmsrq(msr, val)
arch/x86/kernel/cpu/resctrl/monitor.c:  rdmsrl(MSR_IA32_QM_CTR, msr_val);
drivers/hv/mshv_vtl_main.c:                             rdmsrl(reg_table[i].msr_addr, *reg64);
drivers/idle/intel_idle.c:      rdmsrl(MSR_PKG_CST_CONFIG_CONTROL, msr_val);
drivers/idle/intel_idle.c:      rdmsrl(MSR_PKG_CST_CONFIG_CONTROL, msr_val);


Thanks!
    Xin

Re: Time to drop wrmsrl()/rdmsrl()?

[Sean Christopherson]

On Mon, Dec 15, 2025, Xin Li wrote:
> Hi Ingo,
> 
> Keeping them around only causes churn now, so probably it is time to drop
> them?

+1, IMO keeping the old APIs is doing far more harm than good at this point.

> https://lore.kernel.org/lkml/aK8fmMGH0rB2LuA9@google.com/
> 
> https://lore.kernel.org/lkml/aR0vK3z1owwM8X8H@intel.com/
> 
> 
> I think all of their uses were changed to wrmsrq()/rdmsrq(), but they are
> coming back, and with the latest Linux 6.19-rc1 tag, I get:
> 
> $git grep -w rdmsrl
> arch/x86/events/amd/uncore.c:   rdmsrl(hwc->event_base, new);
> arch/x86/include/asm/msr.h:#define rdmsrl(msr, val) rdmsrq(msr, val)
> arch/x86/kernel/cpu/resctrl/monitor.c:  rdmsrl(MSR_IA32_QM_CTR, msr_val);
> drivers/hv/mshv_vtl_main.c:                             rdmsrl(reg_table[i].msr_addr, *reg64);
> drivers/idle/intel_idle.c:      rdmsrl(MSR_PKG_CST_CONFIG_CONTROL, msr_val);
> drivers/idle/intel_idle.c:      rdmsrl(MSR_PKG_CST_CONFIG_CONTROL, msr_val);
> 
> 
> Thanks!
>     Xin