aarch64: push the set of rules before falling into slow path

  It is supposed to save the rules for the instructions before falling into slow path.

Tested in glibc-2.28 before fixing:
Thread 2 "xxxxxxx" hit Breakpoint 1, _dl_tlsdesc_dynamic () at ../sysdeps/aarch64/dl-tlsdesc.S:149
149		stp	x1,  x2, [sp, #-32]!
Missing separate debuginfos, use: dnf debuginfo-install libgcc-7.3.0-20190804.h24.aarch64
(gdb) ni
_dl_tlsdesc_dynamic () at ../sysdeps/aarch64/dl-tlsdesc.S:150
150		stp	x3,  x4, [sp, #16]
(gdb) 
_dl_tlsdesc_dynamic () at ../sysdeps/aarch64/dl-tlsdesc.S:157
157		mrs	x4, tpidr_el0
(gdb) 
158		ldr	PTR_REG (1), [x0,#TLSDESC_ARG]
(gdb) 
159		ldr	PTR_REG (0), [x4,#TCBHEAD_DTV]
(gdb) 
160		ldr	PTR_REG (3), [x1,#TLSDESC_GEN_COUNT]
(gdb) 
161		ldr	PTR_REG (2), [x0,#DTV_COUNTER]
(gdb) 
162		cmp	PTR_REG (3), PTR_REG (2)
(gdb) 
163		b.hi	2f
(gdb) 
165		ldp	PTR_REG (2), PTR_REG (3), [x1,#TLSDESC_MODID]
(gdb) 
166		add	PTR_REG (0), PTR_REG (0), PTR_REG (2), lsl #(PTR_LOG_SIZE + 1)
(gdb) 
167		ldr	PTR_REG (0), [x0] /* Load val member of DTV entry.  */
(gdb) 
168		cmp	PTR_REG (0), #TLS_DTV_UNALLOCATED
(gdb) 
169		b.eq	2f
(gdb) bt
#0  _dl_tlsdesc_dynamic () at ../sysdeps/aarch64/dl-tlsdesc.S:169
#1  0x0000ffffbe4fbb44 in OurFunction (threadId=4294967295)
    at /home/test/test_function.c:30
#2  0x0000000000400c08 in initaaa () at thread.c:58
#3  0x0000000000400c50 in thread_proc (param=0x0) at thread.c:71
#4  0x0000ffffbf6918bc in start_thread (arg=0xfffffffff29f) at pthread_create.c:486
#5  0x0000ffffbf5669ec in thread_start () at ../sysdeps/unix/sysv/linux/aarch64/clone.S:78
(gdb) ni
_dl_tlsdesc_dynamic () at ../sysdeps/aarch64/dl-tlsdesc.S:184
184		stp	x29, x30, [sp,#-16*NSAVEXREGPAIRS]!
(gdb) bt
#0  _dl_tlsdesc_dynamic () at ../sysdeps/aarch64/dl-tlsdesc.S:184
#1  0x0000ffffbe4fbb44 in OurFunction (threadId=4294967295)
    at /home/test/test_function.c:30
#2  0x0000000000000000 in ?? ()
Backtrace stopped: previous frame identical to this frame (corrupt stack?)

Signed-off-by: liqingqing <liqingqing3@huawei.com> 
Signed-off-by: Shuo Wang <wangshuo47@huawei.com>
---
 sysdeps/aarch64/dl-tlsdesc.S | 2 ++
 1 file changed, 2 insertions(+)
  

The 01/04/2021 22:47, Shuo Wang wrote:
> It is supposed to save the rules for the instructions before falling into slow path.
...
> (gdb) bt
> #0  _dl_tlsdesc_dynamic () at ../sysdeps/aarch64/dl-tlsdesc.S:169
> #1  0x0000ffffbe4fbb44 in OurFunction (threadId=4294967295)
>     at /home/test/test_function.c:30
> #2  0x0000000000400c08 in initaaa () at thread.c:58
> #3  0x0000000000400c50 in thread_proc (param=0x0) at thread.c:71
> #4  0x0000ffffbf6918bc in start_thread (arg=0xfffffffff29f) at pthread_create.c:486
> #5  0x0000ffffbf5669ec in thread_start () at ../sysdeps/unix/sysv/linux/aarch64/clone.S:78
> (gdb) ni
> _dl_tlsdesc_dynamic () at ../sysdeps/aarch64/dl-tlsdesc.S:184
> 184		stp	x29, x30, [sp,#-16*NSAVEXREGPAIRS]!
> (gdb) bt
> #0  _dl_tlsdesc_dynamic () at ../sysdeps/aarch64/dl-tlsdesc.S:184
> #1  0x0000ffffbe4fbb44 in OurFunction (threadId=4294967295)
>     at /home/test/test_function.c:30
> #2  0x0000000000000000 in ?? ()
> Backtrace stopped: previous frame identical to this frame (corrupt stack?)
> 
> Signed-off-by: liqingqing <liqingqing3@huawei.com> 
> Signed-off-by: Shuo Wang <wangshuo47@huawei.com>
> ---
>  sysdeps/aarch64/dl-tlsdesc.S | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/sysdeps/aarch64/dl-tlsdesc.S b/sysdeps/aarch64/dl-tlsdesc.S
> index 425e45a50f..aee63cc561 100644
> --- a/sysdeps/aarch64/dl-tlsdesc.S
> +++ b/sysdeps/aarch64/dl-tlsdesc.S
> @@ -170,6 +170,7 @@ _dl_tlsdesc_dynamic:
>  	ldr	PTR_REG (0), [x0] /* Load val member of DTV entry.  */
>  	cmp	PTR_REG (0), #TLS_DTV_UNALLOCATED
>  	b.eq	2f
> +	cfi_remember_state
>  	sub	PTR_REG (3), PTR_REG (3), PTR_REG (4)
>  	add	PTR_REG (0), PTR_REG (0), PTR_REG (3)
>  1:
> @@ -188,6 +189,7 @@ _dl_tlsdesc_dynamic:
>  	cfi_window_save
>  # endif
>  # define NSAVEXREGPAIRS 8
> +	cfi_restore_state

this should be before the cfi_window_save otherwise
the PAC signing state is restored which we dont want.

just move this cfi above the PAC bits.

the patch looks good otherwise.

note that we dont use Signed-off-by: in glibc (which
linux uses for copyright purposes). if you have
coauthors then use Co-authored-by:

>  	stp	x29, x30, [sp,#-16*NSAVEXREGPAIRS]!
>  	cfi_adjust_cfa_offset (16*NSAVEXREGPAIRS)
>  	cfi_rel_offset (x29, 0)
> -- 
> 2.23.0
>