[PATCH 0/1] MIPS: fix optimised memset returning garbage

[PATCH 0/1] MIPS: fix optimised memset returning garbage

[Denis Orlov]

There is currently a bug in memset implementation from
arch/mips/lib/memset.S that is used when MIPS_OPTIMIZED_STRING_FUNCTIONS
option is enabled. It stems from a missing '.set noreorder' directive
before the first branch. This results in an assembler stuffing a 'nop'
instruction in a delay slot instead of an instruction writing the return
value into v0 register that was supposed to go there. That leads to an
error that appears when reading memory from SPI Flash and that was
mentioned here:
http://lists.infradead.org/pipermail/barebox/2021-June/036557.html.

There are a few ways of fixing this bug. The directive '.set noreorder'
that shows up a bit later in code could be moved up and placed right
before the memset label (as it was originally in Linux codebase - see
a583158c9ce822c96a718fbf877cec1e5f9ad75d). Or, rather, the move and the
branch instruction could be swapped, thus allowing the assembler to
properly place move in delay slot (as it is done in Linux now - see
68dec269ee29c3abfd09596fbee7e40d875a6ab3) - that is what is done in
patch.

Denis Orlov (1):
  MIPS: fix optimised memset returning garbage

 arch/mips/lib/memset.S | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

-- 
2.20.1


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

[PATCH 1/1] MIPS: fix optimised memset returning garbage

[Denis Orlov]

There is a subtle bug in arch/mips/lib/memset.S, which results in
not writing the correct return value into v0 register if the value
to fill memory with is 0.

Signed-off-by: Denis Orlov <denorl2009@gmail.com>
---
 arch/mips/lib/memset.S | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/mips/lib/memset.S b/arch/mips/lib/memset.S
index 0b81bd7d8d..cd4e4ccfe8 100644
--- a/arch/mips/lib/memset.S
+++ b/arch/mips/lib/memset.S
@@ -34,8 +34,8 @@
  */
 
 LEAF(memset)
-	beqz		a1, 1f
 	move		v0, a0			/* result */
+	beqz		a1, 1f
 
 	andi		a1, 0xff		/* spread fillword */
 	LONG_SLL		t1, a1, 8
-- 
2.20.1


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

Re: [PATCH 0/1] MIPS: fix optimised memset returning garbage

[Antony Pavlov]

On Thu, 10 Mar 2022 16:40:39 +0300
Denis Orlov <denorl2009@gmail.com> wrote:

Hi Denis!

I'm very glad to see that you have found this anoying bug!

Some tips:

* you comment from "PATCH 0/1" is very informative but after applying "PATCH 1/1"
to the mainline this very informative comment remain in maillist archive but not
in the barebox git repo; Could you please combine "PATCH 0/1" comment and "PATCH 1/1"
comment and resend single "PATCH v2" ?

* I propose this ammendment for your comment:

> --- a/PATCH-0/1
> +++ b/PATCH-0/1
> @@ -108,8 +108,8 @@ List-Archive: <https://lore.barebox.org/barebox/>
>  There is currently a bug in memset implementation from
>  arch/mips/lib/memset.S that is used when MIPS_OPTIMIZED_STRING_FUNCTIONS
>  option is enabled. It stems from a missing '.set noreorder' directive
> -before the first branch. This results in an assembler stuffing a 'nop'
> -instruction in a delay slot instead of an instruction writing the return
> +before the first beqz instruction. This results in an assembler stuffing a 'nop'
> +instruction in the beqz delay slot instead of an instruction writing the return
>  value into v0 register that was supposed to go there. That leads to an
>  error that appears when reading memory from SPI Flash and that was
>  mentioned here:

-- 
Best regards,
  Antony Pavlov



> There is currently a bug in memset implementation from
> arch/mips/lib/memset.S that is used when MIPS_OPTIMIZED_STRING_FUNCTIONS
> option is enabled. It stems from a missing '.set noreorder' directive
> before the first branch. This results in an assembler stuffing a 'nop'
> instruction in a delay slot instead of an instruction writing the return
> value into v0 register that was supposed to go there. That leads to an
> error that appears when reading memory from SPI Flash and that was
> mentioned here:
> http://lists.infradead.org/pipermail/barebox/2021-June/036557.html.
> 
> There are a few ways of fixing this bug. The directive '.set noreorder'
> that shows up a bit later in code could be moved up and placed right
> before the memset label (as it was originally in Linux codebase - see
> a583158c9ce822c96a718fbf877cec1e5f9ad75d). Or, rather, the move and the
> branch instruction could be swapped, thus allowing the assembler to
> properly place move in delay slot (as it is done in Linux now - see
> 68dec269ee29c3abfd09596fbee7e40d875a6ab3) - that is what is done in
> patch.
> 
> Denis Orlov (1):
>   MIPS: fix optimised memset returning garbage
> 
>  arch/mips/lib/memset.S | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> -- 
> 2.20.1
> 
> 
> _______________________________________________
> barebox mailing list
> barebox@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/barebox


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

[PATCH v2] MIPS: fix optimised memset returning garbage

[Denis Orlov]

There is a subtle bug in arch/mips/lib/memset.S, used if option
MIPS_OPTIMIZED_STRING_FUNCTIONS is enabled, which results in not writing
the correct return value into v0 register if the value to fill memory
with is 0.

The bug stems from a missing '.set noreorder' directive before the first
beqz instruction. This results in an assembler stuffing a 'nop'
instruction in the beqz delay slot instead of an instruction writing the
return value into v0 register that was supposed to go there. That leads
to an error that appears when reading memory from SPI Flash and that was
mentioned here:
http://lists.infradead.org/pipermail/barebox/2021-June/036557.html

There are a few ways of fixing this bug. The directive '.set noreorder'
that shows up a bit later in code could be moved up and placed right
before the memset label (as it was originally in Linux codebase - see
a583158c9ce822c96a718fbf877cec1e5f9ad75d). Or, rather, the move and the
branch instruction could be swapped, thus allowing the assembler to
properly place move in delay slot (as it is done in Linux now - see
68dec269ee29c3abfd09596fbee7e40d875a6ab3) - that is what is done in this
patch.

Signed-off-by: Denis Orlov <denorl2009@gmail.com>
---
 arch/mips/lib/memset.S | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/mips/lib/memset.S b/arch/mips/lib/memset.S
index 0b81bd7d8d..cd4e4ccfe8 100644
--- a/arch/mips/lib/memset.S
+++ b/arch/mips/lib/memset.S
@@ -34,8 +34,8 @@
  */
 
 LEAF(memset)
-	beqz		a1, 1f
 	move		v0, a0			/* result */
+	beqz		a1, 1f
 
 	andi		a1, 0xff		/* spread fillword */
 	LONG_SLL		t1, a1, 8
-- 
2.20.1


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

Re: [PATCH v2] MIPS: fix optimised memset returning garbage

[Sascha Hauer]

On Thu, Mar 10, 2022 at 11:42:06PM +0300, Denis Orlov wrote:
> There is a subtle bug in arch/mips/lib/memset.S, used if option
> MIPS_OPTIMIZED_STRING_FUNCTIONS is enabled, which results in not writing
> the correct return value into v0 register if the value to fill memory
> with is 0.
> 
> The bug stems from a missing '.set noreorder' directive before the first
> beqz instruction. This results in an assembler stuffing a 'nop'
> instruction in the beqz delay slot instead of an instruction writing the
> return value into v0 register that was supposed to go there. That leads
> to an error that appears when reading memory from SPI Flash and that was
> mentioned here:
> http://lists.infradead.org/pipermail/barebox/2021-June/036557.html
> 
> There are a few ways of fixing this bug. The directive '.set noreorder'
> that shows up a bit later in code could be moved up and placed right
> before the memset label (as it was originally in Linux codebase - see
> a583158c9ce822c96a718fbf877cec1e5f9ad75d). Or, rather, the move and the
> branch instruction could be swapped, thus allowing the assembler to
> properly place move in delay slot (as it is done in Linux now - see
> 68dec269ee29c3abfd09596fbee7e40d875a6ab3) - that is what is done in this
> patch.
> 
> Signed-off-by: Denis Orlov <denorl2009@gmail.com>
> ---
>  arch/mips/lib/memset.S | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)

Applied, thanks

Sascha

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox