2024-04-28 17:31:28

by Jules Irenge

Subject: [PATCH 1/3] x86/tsc: Use div64_ul instead of do_div()

do_div() truncates a u64 divisor to 32 bit.
This can lead to non-zero being truncated to zero for division.

Fix coccinelle warning
WARNING: do_div() does a 64-by-32 division, please consider using div64_ul instead

Signed-off-by: Jules Irenge <[email protected]>
---
arch/x86/kernel/tsc.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
index 5a69a49acc96..2da37f33dd17 100644
--- a/arch/x86/kernel/tsc.c
+++ b/arch/x86/kernel/tsc.c
@@ -484,7 +484,7 @@ static unsigned long pit_calibrate_tsc(u32 latch, unsigned long ms, int loopmin)

/* Calculate the PIT value */
delta = t2 - t1;
- do_div(delta, ms);
+ div64_ul(delta, ms);
return delta;
}

--
2.43.2

2024-04-28 19:14:36

by H. Peter Anvin

Subject: Re: [PATCH 1/3] x86/tsc: Use div64_ul instead of do_div()

On April 28, 2024 10:31:16 AM PDT, Jules Irenge <[email protected]> wrote:
>do_div() truncates a u64 divisor to 32 bit.
>This can lead to non-zero being truncated to zero for division.
>
>Fix coccinelle warning
>WARNING: do_div() does a 64-by-32 division, please consider using div64_ul instead
>
>Signed-off-by: Jules Irenge <[email protected]>
>---
> arch/x86/kernel/tsc.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
>diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
>index 5a69a49acc96..2da37f33dd17 100644
>--- a/arch/x86/kernel/tsc.c
>+++ b/arch/x86/kernel/tsc.c
>@@ -484,7 +484,7 @@ static unsigned long pit_calibrate_tsc(u32 latch, unsigned long ms, int loopmin)
>
> /* Calculate the PIT value */
> delta = t2 - t1;
>- do_div(delta, ms);
>+ div64_ul(delta, ms);
> return delta;
> }
>

Are you sure this is not a false positive? This is a *much* more expensive operation on 32 bits.

2024-05-02 16:31:21

by David Laight

Subject: RE: [PATCH 1/3] x86/tsc: Use div64_ul instead of do_div()

From: H. Peter Anvin
> Sent: 28 April 2024 20:13
>
> On April 28, 2024 10:31:16 AM PDT, Jules Irenge <[email protected]> wrote:
> >do_div() truncates a u64 divisor to 32 bit.
> >This can lead to non-zero being truncated to zero for division.
> >
> >Fix coccinelle warning
> >WARNING: do_div() does a 64-by-32 division, please consider using div64_ul instead
> >
> >Signed-off-by: Jules Irenge <[email protected]>
> >---
> > arch/x86/kernel/tsc.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> >diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
> >index 5a69a49acc96..2da37f33dd17 100644
> >--- a/arch/x86/kernel/tsc.c
> >+++ b/arch/x86/kernel/tsc.c
> >@@ -484,7 +484,7 @@ static unsigned long pit_calibrate_tsc(u32 latch, unsigned long ms, int loopmin)
> >
> > /* Calculate the PIT value */
> > delta = t2 - t1;
> >- do_div(delta, ms);
> >+ div64_ul(delta, ms);
> > return delta;
> > }
> >
>
> Are you sure this is not a false positive? This is a *much* more expensive operation on 32 bits.

IIRC do_div() isn't that great either.
Somewhere it changed from being a simple wrapper on the x86 'div' instruction
(ie 64 by 32 giving 32 and 32) to being able to generate a 64bit quotent.

But the patch is entirely broken as well.

David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)