Re: [PATCH v7 15/15] qla2xxx: Fix endianness annotations in source files

From: Bart Van Assche <bvanassche@acm.org>
To: Finn Thain <fthain@telegraphics.com.au>, Daniel Wagner <dwagner@suse.de>
Cc: "Martin K. Petersen" <martin.petersen@oracle.com>,
	"James E. J. Bottomley" <jejb@linux.vnet.ibm.com>,
	linux-scsi@vger.kernel.org, Hannes Reinecke <hare@suse.de>,
	Arun Easi <aeasi@marvell.com>,
	Himanshu Madhani <himanshu.madhani@oracle.com>,
	Nilesh Javali <njavali@marvell.com>,
	Quinn Tran <qutran@marvell.com>, Martin Wilck <mwilck@suse.com>,
	Roman Bolshakov <r.bolshakov@yadro.com>
Subject: Re: [PATCH v7 15/15] qla2xxx: Fix endianness annotations in source files
Date: Sun, 24 May 2020 08:50:47 -0700	[thread overview]
Message-ID: <2737709e-1423-da29-4697-dc517ad7e413@acm.org> (raw)
In-Reply-To: <alpine.LNX.2.22.394.2005211358460.8@nippy.intranet>

On 2020-05-23 21:28, Finn Thain wrote:
> 2. The get_unaligned_le32() changes produce new pointer offsets in the 
> assembly code for qla82xx_get_table_desc() and qla82xx_get_data_desc().
> 
> diff -ru /tmp/unpatched/qla_target.s /tmp/patched/qla_target.s
> --- /tmp/unpatched/qla_target.s 2020-05-24 14:02:32.178019380 +1000
> +++ /tmp/patched/qla_target.s   2020-05-24 14:01:43.487947966 +1000
> @@ -12884,10 +12884,10 @@
>         .cfi_offset 6, -16
>         movq    %rsp, %rbp
>         .cfi_def_cfa_register 6
> -       subq    $32, %rsp
> -       movq    %rdi, -24(%rbp)
> -       movq    %rsi, -32(%rbp)
> -       movq    -32(%rbp), %rax
> +       subq    $64, %rsp
> +       movq    %rdi, -56(%rbp)
> +       movq    %rsi, -64(%rbp)
> +       movq    -64(%rbp), %rax
>         movl    52(%rax), %eax
>         movl    %eax, -8(%rbp)
>         movl    $24, -12(%rbp)
> @@ -12895,62 +12895,62 @@
>         cmpl    %eax, -12(%rbp)
>         cmovbe  -12(%rbp), %eax
>         movl    %eax, %edx
> -       movq    -32(%rbp), %rax
> +       movq    -64(%rbp), %rax
>         movl    %edx, 52(%rax)
> -       movq    -24(%rbp), %rax
> +       movq    -56(%rbp), %rax
> (and so on.)
> 
> Was this expected? I find it surprising...

The functions qla82xx_get_table_desc() and qla82xx_get_data_desc() exist
in qla_nx.c. Does the above diff perhaps refer to qla_nx.s instead of
qla_target.s?

To me the change of "subq $32, %rsp" into "subq $64, %rsp" means that
the compiler reserved more space on the stack for local variables.

If I compare the assembler output for qla_nx.c then I see that
qla82xx_get_table_desc() gets inlined with my patch applied but not
without my patch applied. This is something that I had not expected but
that explains the above diff IMHO.

Bart.