Re: [PATCH 0/5] x86: Cleanups around slow_down_io()

["Jürgen Groß" <jgross@suse.com>]

[-- Attachment #1.1.1: Type: text/plain, Size: 2686 bytes --]

On 14.12.25 09:05, Ingo Molnar wrote:
> 
> * Juergen Gross <jgross@suse.com> wrote:
> 
>> While looking at paravirt cleanups I stumbled over slow_down_io() and
>> the related REALLY_SLOW_IO define.
>>
>> Especially REALLY_SLOW_IO is a mess, which is proven by 2 completely
>> wrong use cases.
>>
>> Do several cleanups, resulting in a deletion of REALLY_SLOW_IO and the
>> io_delay() paravirt function hook.
>>
>> Patches 2 and 3 are not changing any functionality, but maybe they
>> should? As the potential bug has been present for more than a decade
>> now, I went with just deleting the useless "#define REALLY_SLOW_IO".
>> The alternative would be to do something similar as in patch 5.
>>
>> Juergen Gross (5):
>>    x86/paravirt: Replace io_delay() hook with a bool
>>    hwmon/lm78: Drop REALLY_SLOW_IO setting
>>    hwmon/w83781d: Drop REALLY_SLOW_IO setting
>>    block/floppy: Don't use REALLY_SLOW_IO for delays
>>    x86/io: Remove REALLY_SLOW_IO handling
>>
>>   arch/x86/include/asm/floppy.h         | 27 ++++++++++++++++++++++-----
>>   arch/x86/include/asm/io.h             | 12 +++++-------
>>   arch/x86/include/asm/paravirt.h       | 11 +----------
>>   arch/x86/include/asm/paravirt_types.h |  3 +--
>>   arch/x86/kernel/cpu/vmware.c          |  2 +-
>>   arch/x86/kernel/kvm.c                 |  8 +-------
>>   arch/x86/kernel/paravirt.c            |  3 +--
>>   arch/x86/xen/enlighten_pv.c           |  6 +-----
>>   drivers/block/floppy.c                |  2 --
>>   drivers/hwmon/lm78.c                  |  5 +++--
>>   drivers/hwmon/w83781d.c               |  5 +++--
>>   11 files changed, 39 insertions(+), 45 deletions(-)
> 
> I think we should get rid of *all* io_delay hacks, they might have been
> relevant in the days of i386 systems, but we don't even support i386
> CPUs anymore. Should it cause any regressions, it's easy to bisect to.
> There's been enough changes around all these facilities that the
> original timings are probably way off already, so we've just been
> cargo-cult porting these to newer kernels essentially.

Fine with me.

Which path to removal of io_delay would you (and others) prefer?

1. Ripping it out immediately.

2. Hiding it behind a default-off config option for a few kernel versions
    before removing it.

3. Using CONFIG_IO_DELAY_NONE as the default io_delay_type before ripping it
    out.

4. Using CONFIG_IO_DELAY_NONE as the default io_delay_type before hiding it
    behind a default-off config option, then rip it out later.

In cases 2-4 I'd still like to have patch 1 of my series applied, as it will
make paravirt rework easier.


Juergen

[-- Attachment #1.1.2: OpenPGP public key --]
[-- Type: application/pgp-keys, Size: 3743 bytes --]

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 495 bytes --]