lkml.org 
[lkml]   [2019]   [Aug]   [20]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: [PATCH v8 11/28] x86/asm/head: annotate data appropriatelly
> Subject: Re: [PATCH v8 11/28] x86/asm/head: annotate data appropriatelly

appropriately

On Thu, Aug 08, 2019 at 12:38:37PM +0200, Jiri Slaby wrote:
> Use the new SYM_DATA, SYM_DATA_START, and SYM_DATA_END in both 32 and 64
> bit heads. In the 64-bit version, define also
> SYM_DATA_START_PAGE_ALIGNED locally using the new SYM_START. It is used
> in the code instead of NEXT_PAGE() which was defined in this file and
> has been using the obsolete macro GLOBAL().
>
> Now, the data in the 64-bit object file look sane:
> Value Size Type Bind Vis Ndx Name
> 0000 4096 OBJECT GLOBAL DEFAULT 15 init_level4_pgt
> 1000 4096 OBJECT GLOBAL DEFAULT 15 level3_kernel_pgt
> 2000 2048 OBJECT GLOBAL DEFAULT 15 level2_kernel_pgt
> 3000 4096 OBJECT GLOBAL DEFAULT 15 level2_fixmap_pgt
> 4000 4096 OBJECT GLOBAL DEFAULT 15 level1_fixmap_pgt
> 5000 2 OBJECT GLOBAL DEFAULT 15 early_gdt_descr
> 5002 8 OBJECT LOCAL DEFAULT 15 early_gdt_descr_base
> 500a 8 OBJECT GLOBAL DEFAULT 15 phys_base
> 0000 8 OBJECT GLOBAL DEFAULT 17 initial_code
> 0008 8 OBJECT GLOBAL DEFAULT 17 initial_gs
> 0010 8 OBJECT GLOBAL DEFAULT 17 initial_stack
> 0000 4 OBJECT GLOBAL DEFAULT 19 early_recursion_flag
> 1000 4096 OBJECT GLOBAL DEFAULT 19 early_level4_pgt
> 2000 0x40000 OBJECT GLOBAL DEFAULT 19 early_dynamic_pgts
> 0000 4096 OBJECT GLOBAL DEFAULT 22 empty_zero_page
>
> All have correct size and type.

Nice.

>
> Note, that we can now see that it might be worth pushing
> early_recursion_flag after early_dynamic_pgts -- we are wasting almost
> 4K of .init.data.

Yes, please do in a separate patch which can even go separately. I get
here:

---
Disassembly of section .init.data:

ffffffff82684000 <early_recursion_flag>:
...

ffffffff82685000 <early_top_pgt>:
...

ffffffff82686000 <early_dynamic_pgts>:
...

ffffffff826c6000 <next_early_pgt>:
...

ffffffff826c6020 <initcall_level_names>:
---


vs


---
Disassembly of section .init.data:

ffffffff82684000 <early_top_pgt>:
...

ffffffff82685000 <early_dynamic_pgts>:
...

ffffffff826c5000 <early_recursion_flag>:
ffffffff826c5000: 00 00 add %al,(%rax)
...

ffffffff826c5004 <next_early_pgt>:
...

ffffffff826c5020 <initcall_level_names>:
---

That's exactly 4K saved.


> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: Ingo Molnar <mingo@redhat.com>
> Cc: "H. Peter Anvin" <hpa@zytor.com>
> Cc: x86@kernel.org
> ---
> arch/x86/kernel/head_32.S | 29 ++++++++-------
> arch/x86/kernel/head_64.S | 78 +++++++++++++++++++++------------------
> 2 files changed, 58 insertions(+), 49 deletions(-)
>
> diff --git a/arch/x86/kernel/head_32.S b/arch/x86/kernel/head_32.S
> index 0bae769b7b59..2d5390d84467 100644
> --- a/arch/x86/kernel/head_32.S
> +++ b/arch/x86/kernel/head_32.S
> @@ -502,8 +502,7 @@ ENDPROC(early_ignore_irq)
>
> __INITDATA
> .align 4
> -GLOBAL(early_recursion_flag)
> - .long 0
> +SYM_DATA(early_recursion_flag, .long 0)
>
> __REFDATA
> .align 4
> @@ -551,7 +550,7 @@ EXPORT_SYMBOL(empty_zero_page)
> __PAGE_ALIGNED_DATA
> /* Page-aligned for the benefit of paravirt? */
> .align PGD_ALIGN
> -ENTRY(initial_page_table)
> +SYM_DATA_START(initial_page_table)
> .long pa(initial_pg_pmd+PGD_IDENT_ATTR),0 /* low identity map */
> # if KPMDS == 3
> .long pa(initial_pg_pmd+PGD_IDENT_ATTR),0
> @@ -569,17 +568,18 @@ ENTRY(initial_page_table)
> # error "Kernel PMDs should be 1, 2 or 3"
> # endif
> .align PAGE_SIZE /* needs to be page-sized too */
> +SYM_DATA_END(initial_page_table)
> #endif
>
> .data
> .balign 4
> -ENTRY(initial_stack)
> - /*
> - * The SIZEOF_PTREGS gap is a convention which helps the in-kernel
> - * unwinder reliably detect the end of the stack.
> - */
> - .long init_thread_union + THREAD_SIZE - SIZEOF_PTREGS - \
> - TOP_OF_KERNEL_STACK_PADDING;
> +/*
> + * The SIZEOF_PTREGS gap is a convention which helps the in-kernel unwinder
> + * reliably detect the end of the stack.
> + */
> +SYM_DATA(initial_stack,
> + .long init_thread_union + THREAD_SIZE -
> + SIZEOF_PTREGS - TOP_OF_KERNEL_STACK_PADDING)
>
> __INITRODATA
> int_msg:
> @@ -600,22 +600,25 @@ int_msg:
> ALIGN
> # early boot GDT descriptor (must use 1:1 address mapping)
> .word 0 # 32 bit align gdt_desc.address
> -boot_gdt_descr:
> +SYM_DATA_START(boot_gdt_descr)
> .word __BOOT_DS+7
> .long boot_gdt - __PAGE_OFFSET
> +SYM_DATA_END(boot_gdt_descr)

So there's one "globl boot_gdt_descr" above already and this turns into:

.data
.globl boot_gdt_descr
^^^^^^^^^^^^^^^^^^^^^

.align 4,0x90
# early boot GDT descriptor (must use 1:1 address mapping)
.word 0 # 32 bit align gdt_desc.address
.globl boot_gdt_descr ; ; boot_gdt_descr:
^^^^^^^^^^^^^^^^^^^^^

I guess you can remove the above one.

Also, this can be made a local symbol too.

> # boot GDT descriptor (later on used by CPU#0):
> .word 0 # 32 bit align gdt_desc.address
> -ENTRY(early_gdt_descr)
> +SYM_DATA_START(early_gdt_descr)
> .word GDT_ENTRIES*8-1
> .long gdt_page /* Overwritten for secondary CPUs */
> +SYM_DATA_END(early_gdt_descr)
>
> /*
> * The boot_gdt must mirror the equivalent in setup.S and is
> * used only for booting.
> */
> .align L1_CACHE_BYTES
> -ENTRY(boot_gdt)
> +SYM_DATA_START(boot_gdt)
> .fill GDT_ENTRY_BOOT_CS,8,0
> .quad 0x00cf9a000000ffff /* kernel 4GB code at 0x00000000 */
> .quad 0x00cf92000000ffff /* kernel 4GB data at 0x00000000 */
> +SYM_DATA_END(boot_gdt)
> diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S
> index 6fedcda37634..6661c76a2049 100644
> --- a/arch/x86/kernel/head_64.S
> +++ b/arch/x86/kernel/head_64.S
> @@ -260,16 +260,14 @@ END(start_cpu0)
> /* Both SMP bootup and ACPI suspend change these variables */
> __REFDATA
> .balign 8
> - GLOBAL(initial_code)
> - .quad x86_64_start_kernel
> - GLOBAL(initial_gs)
> - .quad INIT_PER_CPU_VAR(fixed_percpu_data)
> - GLOBAL(initial_stack)
> - /*
> - * The SIZEOF_PTREGS gap is a convention which helps the in-kernel
> - * unwinder reliably detect the end of the stack.
> - */
> - .quad init_thread_union + THREAD_SIZE - SIZEOF_PTREGS
> +SYM_DATA(initial_code, .quad x86_64_start_kernel)
> +SYM_DATA(initial_gs, .quad INIT_PER_CPU_VAR(fixed_percpu_data))

<---- newline here.

> +/*
> + * The SIZEOF_PTREGS gap is a convention which helps the in-kernel unwinder
> + * reliably detect the end of the stack.
> + */
> +SYM_DATA(initial_stack,
> + .quad init_thread_union + THREAD_SIZE - SIZEOF_PTREGS)

No need to break that line.

...

--
Regards/Gruss,
Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.

\
 
 \ /
  Last update: 2019-08-20 18:27    [W:0.788 / U:0.128 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site