[PATCH] blk-iocost: don't make all constants unsigned long long

[PATCH] blk-iocost: don't make all constants unsigned long long

[Yann Droneaud]

My shiny new compiler (GCC 13) is reporting the following
warnings:

  ../block/blk-iocost.c: In function 'ioc_weight_prfill':
  ../block/blk-iocost.c:3035:37: warning: format '%u' expects argument of type 'unsigned int', but argument 4 has type 'long unsigned int' [-Wformat=]
   3035 |                 seq_printf(sf, "%s %u\n", dname, iocg->cfg_weight / WEIGHT_ONE);
        |                                    ~^            ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
        |                                     |                             |
        |                                     unsigned int                  long unsigned int
        |                                    %lu
  ../block/blk-iocost.c: In function 'ioc_weight_show':
  ../block/blk-iocost.c:3045:34: warning: format '%u' expects argument of type 'unsigned int', but argument 3 has type 'long unsigned int' [-Wformat=]
   3045 |         seq_printf(sf, "default %u\n", iocc->dfl_weight / WEIGHT_ONE);
        |                                 ~^     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
        |                                  |                      |
        |                                  unsigned int           long unsigned int
        |                                 %lu

It appears WEIGHT_ONE enum is unnecessarly unsigned long
(or unsigned long long on 32bit) because of VTIME_PER_SEC
and/or AUTOP_CYCLE_NSEC need the enum to be that large.

Addressed by lazy splitting the "catch all" anonymous
enum and placing the unsigned long long constants in
their own anonymous enums.

Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
---
 block/blk-iocost.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/block/blk-iocost.c b/block/blk-iocost.c
index 495396425bad..bb1f8522c0f1 100644
--- a/block/blk-iocost.c
+++ b/block/blk-iocost.c
@@ -232,7 +232,9 @@ enum {
 
 	/* 1/64k is granular enough and can easily be handled w/ u32 */
 	WEIGHT_ONE		= 1 << 16,
+};
 
+enum {
 	/*
 	 * As vtime is used to calculate the cost of each IO, it needs to
 	 * be fairly high precision.  For example, it should be able to
@@ -255,7 +257,9 @@ enum {
 
 	VRATE_MIN		= VTIME_PER_USEC * VRATE_MIN_PPM / MILLION,
 	VRATE_CLAMP_ADJ_PCT	= 4,
+};
 
+enum {
 	/* if IOs end up waiting for requests, issue less */
 	RQ_WAIT_BUSY_PCT	= 5,
 
@@ -293,10 +297,14 @@ enum {
 
 	/* don't let cmds which take a very long time pin lagging for too long */
 	MAX_LAGGING_PERIODS	= 10,
+};
 
+enum {
 	/* switch iff the conditions are met for longer than this */
 	AUTOP_CYCLE_NSEC	= 10LLU * NSEC_PER_SEC,
+};
 
+enum {
 	/*
 	 * Count IO size in 4k pages.  The 12bit shift helps keeping
 	 * size-proportional components of cost calculation in closer
-- 
2.37.2

Re: [PATCH] blk-iocost: don't make all constants unsigned long long

[Michal Koutný]

[-- Attachment #1: Type: text/plain, Size: 338 bytes --]

On Tue, Dec 20, 2022 at 09:18:19PM +0100, Yann Droneaud <ydroneaud@opteya.com> wrote:
> +enum {
>  	/* switch iff the conditions are met for longer than this */
>  	AUTOP_CYCLE_NSEC	= 10LLU * NSEC_PER_SEC,
> +};

This looks gratuitous.

What about indivudial #defines with typed literals instead of the "lazy
splitting"?

Regards,
Michal

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

Re: [PATCH] blk-iocost: don't make all constants unsigned long long

[Tejun Heo]

On Thu, Dec 22, 2022 at 02:58:55PM +0100, Michal Koutný wrote:
> On Tue, Dec 20, 2022 at 09:18:19PM +0100, Yann Droneaud <ydroneaud@opteya.com> wrote:
> > +enum {
> >  	/* switch iff the conditions are met for longer than this */
> >  	AUTOP_CYCLE_NSEC	= 10LLU * NSEC_PER_SEC,
> > +};
> 
> This looks gratuitous.
> 
> What about indivudial #defines with typed literals instead of the "lazy
> splitting"?

enums are so much better for debugging and tracing. This is a gcc caused
problem where there's no other way to generate the same code between two gcc
versions without splitting the enum definitions. I'm kinda baffled that this
is what they chose to do but can't think of a better way to work around it.

Thanks.

-- 
tejun

Re: [PATCH] blk-iocost: don't make all constants unsigned long long

[Tejun Heo]

On Wed, Jan 04, 2023 at 12:15:06PM -1000, Tejun Heo wrote:
> On Thu, Dec 22, 2022 at 02:58:55PM +0100, Michal Koutný wrote:
> > On Tue, Dec 20, 2022 at 09:18:19PM +0100, Yann Droneaud <ydroneaud@opteya.com> wrote:
> > > +enum {
> > >  	/* switch iff the conditions are met for longer than this */
> > >  	AUTOP_CYCLE_NSEC	= 10LLU * NSEC_PER_SEC,
> > > +};
> > 
> > This looks gratuitous.
> > 
> > What about indivudial #defines with typed literals instead of the "lazy
> > splitting"?
> 
> enums are so much better for debugging and tracing. This is a gcc caused
> problem where there's no other way to generate the same code between two gcc
> versions without splitting the enum definitions. I'm kinda baffled that this
> is what they chose to do but can't think of a better way to work around it.

I thought this was the other patch addressing this issue. The proposed patch
is rather painful to look at. The other one splits it into two groups.

Thanks.

-- 
tejun

Re: [PATCH] blk-iocost: don't make all constants unsigned long long

[Tejun Heo]

On Tue, Dec 20, 2022 at 09:18:19PM +0100, Yann Droneaud wrote:
> My shiny new compiler (GCC 13) is reporting the following
> warnings:
> 
>   ../block/blk-iocost.c: In function 'ioc_weight_prfill':
>   ../block/blk-iocost.c:3035:37: warning: format '%u' expects argument of type 'unsigned int', but argument 4 has type 'long unsigned int' [-Wformat=]
>    3035 |                 seq_printf(sf, "%s %u\n", dname, iocg->cfg_weight / WEIGHT_ONE);
>         |                                    ~^            ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>         |                                     |                             |
>         |                                     unsigned int                  long unsigned int
>         |                                    %lu
>   ../block/blk-iocost.c: In function 'ioc_weight_show':
>   ../block/blk-iocost.c:3045:34: warning: format '%u' expects argument of type 'unsigned int', but argument 3 has type 'long unsigned int' [-Wformat=]
>    3045 |         seq_printf(sf, "default %u\n", iocc->dfl_weight / WEIGHT_ONE);
>         |                                 ~^     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>         |                                  |                      |
>         |                                  unsigned int           long unsigned int
>         |                                 %lu
> 
> It appears WEIGHT_ONE enum is unnecessarly unsigned long
> (or unsigned long long on 32bit) because of VTIME_PER_SEC
> and/or AUTOP_CYCLE_NSEC need the enum to be that large.
> 
> Addressed by lazy splitting the "catch all" anonymous
> enum and placing the unsigned long long constants in
> their own anonymous enums.
> 
> Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>

There's a better patch doing this which groups the enums into two groups.
Let's do that instead.

Thanks.

-- 
tejun