Re: [PATCH 1/3] x86: get rid of KERNEL_STACK_OFFSET

[Denys Vlasenko <dvlasenk@redhat.com>]

On 03/18/2015 09:54 PM, Andy Lutomirski wrote:
> On Wed, Mar 18, 2015 at 12:47 PM, Denys Vlasenko <dvlasenk@redhat.com> wrote:
>> PER_CPU_VAR(kernel_stack) was set up in a way where it points
>> five stack slots below the top of stack.
>>
>> Presumably, it was done to avoid one "sub $5*8,%rsp"
>> in syscall/sysenter code paths, where iret frame needs to be
>> created by hand.
>>
>> Ironically, none of them benefits from this optimization,
>> since all of them need to allocate additional data on stack
>> (struct pt_regs), so they still have to perform subtraction.
>>
>> This patch eliminates KERNEL_STACK_OFFSET.
>>
>> PER_CPU_VAR(kernel_stack) now points directly to top of stack.
>> pt_regs allocations are adjusted to allocate iret frame as well.
>> Hopefully we can merge it later with 32-bit specific
>> PER_CPU_VAR(cpu_current_top_of_stack) variable...
>>
>> Semi-mysterious expressions THREAD_INFO(%rsp,RIP) - "why RIP??"
>> are now replaced by more logical THREAD_INFO(%rsp,SIZEOF_PTREGS) -
>> "calculate thread_info's address using information that
>> rsp is SIZEOF_PTREGS bytes below the stack top".
>>
>> Net result in generated code is that constants in several insns
>> are changed.
>>
>> This change is necessary for changing struct pt_regs creation
>> in SYSCALL64 code path from MOV to PUSH instructions.
>>
> 
> Would it be reasonable to break this up into two pieces: first, remove
> KERNEL_STACK_OFFSET from THREAD_INFO and related macros (i.e. make
> them relative to current_top_of_stack instead)

PER_CPU(cpu_current_top_of_stack) exists only in 32 bits.
Can't use it in 64-bit code.

PER_CPU(cpu_current_top_of_stack) becomes equal
to PER_CPU(kernel_stack) only after this patch.

I plan to clean up PER_CPU(cpu_current_top_of_stack)/
PER_CPU(kernel_stack)/PER_CPU(tss->sp0) mess
on top of my patches.

First: this seems to be the easier way.

Second (why do this *after* this patch set, not in it?):
the removal of KERNEL_STACK_OFFSET has no strong reason behind it
unless we do MOV -> PUSH conversion - in which case it is mandatory:
we must have RSP start at top of stack in 64-bit mode.

I want to post a minimal patch set which results in PUSH conversion.
Want to avoid polluting it with tangential goal of removing
duplicated percpu variable.