| Message ID | 1578546590-24737-1-git-send-email-spujar@nvidia.com |
|---|---|
| State | Changes Requested |
| Headers | show |
| Series | [v3] regmap: add iopoll-like atomic polling macro | expand |
09.01.2020 08:09, Sameer Pujar пишет: > This patch adds a macro 'regmap_read_poll_timeout_atomic' that works > similar to 'readx_poll_timeout_atomic' defined in linux/iopoll.h; This > is atomic version of already available 'regmap_read_poll_timeout' macro. > > It should be noted that above atomic macro cannot be used by all regmaps. > If the regmap is set up for atomic use (flat or no cache and MMIO) then > only it can use. > > Signed-off-by: Sameer Pujar <spujar@nvidia.com> > --- Could you please explain what is the targeted use-case here?
On 1/9/2020 11:30 AM, Dmitry Osipenko wrote: > External email: Use caution opening links or attachments > > > 09.01.2020 08:09, Sameer Pujar пишет: >> This patch adds a macro 'regmap_read_poll_timeout_atomic' that works >> similar to 'readx_poll_timeout_atomic' defined in linux/iopoll.h; This >> is atomic version of already available 'regmap_read_poll_timeout' macro. >> >> It should be noted that above atomic macro cannot be used by all regmaps. >> If the regmap is set up for atomic use (flat or no cache and MMIO) then >> only it can use. >> >> Signed-off-by: Sameer Pujar <spujar@nvidia.com> >> --- > Could you please explain what is the targeted use-case here? I was trying to use regmap_read_poll_timeout() to poll for status change of a register. This resulted in "BUG: scheduling while atomic". The callback function, in which I was trying to use the macro, runs in atomic context. Hence new atomic macro is added. I was checking ALSA playback/capture and trigger() callback had to monitor some register status. In general, the new macro can be used in atomic callbacks where regmap interface is used and polling is required.
09.01.2020 10:24, Sameer Pujar пишет: > > On 1/9/2020 11:30 AM, Dmitry Osipenko wrote: >> External email: Use caution opening links or attachments >> >> >> 09.01.2020 08:09, Sameer Pujar пишет: >>> This patch adds a macro 'regmap_read_poll_timeout_atomic' that works >>> similar to 'readx_poll_timeout_atomic' defined in linux/iopoll.h; This >>> is atomic version of already available 'regmap_read_poll_timeout' macro. >>> >>> It should be noted that above atomic macro cannot be used by all >>> regmaps. >>> If the regmap is set up for atomic use (flat or no cache and MMIO) then >>> only it can use. >>> >>> Signed-off-by: Sameer Pujar <spujar@nvidia.com> >>> --- >> Could you please explain what is the targeted use-case here? > > I was trying to use regmap_read_poll_timeout() to poll for status change > of a register. This resulted in "BUG: scheduling while atomic". The > callback function, in which I was trying to use the macro, runs in > atomic context. Hence new atomic macro is added. I was checking ALSA > playback/capture and trigger() callback had to monitor some register > status. > > In general, the new macro can be used in atomic callbacks where regmap > interface is used and polling is required. > You should send a full patchset because it may turn out that the patch which makes use of the new feature isn't correct or maybe the new feature isn't really needed. If there was a previous discussion about the need for this change, then you should provide a link to that discussion. Please note that usually changes without a real use-case in kernel are not getting picked up or they are getting removed later on if nobody makes use of them, so I assume this is a kind of an RFC patch for now.
On 1/9/2020 7:27 PM, Dmitry Osipenko wrote: > External email: Use caution opening links or attachments > > > 09.01.2020 10:24, Sameer Pujar пишет: >> On 1/9/2020 11:30 AM, Dmitry Osipenko wrote: >>> External email: Use caution opening links or attachments >>> >>> >>> 09.01.2020 08:09, Sameer Pujar пишет: >>>> This patch adds a macro 'regmap_read_poll_timeout_atomic' that works >>>> similar to 'readx_poll_timeout_atomic' defined in linux/iopoll.h; This >>>> is atomic version of already available 'regmap_read_poll_timeout' macro. >>>> >>>> It should be noted that above atomic macro cannot be used by all >>>> regmaps. >>>> If the regmap is set up for atomic use (flat or no cache and MMIO) then >>>> only it can use. >>>> >>>> Signed-off-by: Sameer Pujar <spujar@nvidia.com> >>>> --- >>> Could you please explain what is the targeted use-case here? >> I was trying to use regmap_read_poll_timeout() to poll for status change >> of a register. This resulted in "BUG: scheduling while atomic". The >> callback function, in which I was trying to use the macro, runs in >> atomic context. Hence new atomic macro is added. I was checking ALSA >> playback/capture and trigger() callback had to monitor some register >> status. >> >> In general, the new macro can be used in atomic callbacks where regmap >> interface is used and polling is required. >> > You should send a full patchset because it may turn out that the patch > which makes use of the new feature isn't correct or maybe the new > feature isn't really needed. > > If there was a previous discussion about the need for this change, then > you should provide a link to that discussion. > > Please note that usually changes without a real use-case in kernel are > not getting picked up or they are getting removed later on if nobody > makes use of them, so I assume this is a kind of an RFC patch for now. OK. I will send this as part of the complete series. Thank you.
10.01.2020 07:50, Sameer Pujar пишет: > > On 1/9/2020 7:27 PM, Dmitry Osipenko wrote: >> External email: Use caution opening links or attachments >> >> >> 09.01.2020 10:24, Sameer Pujar пишет: >>> On 1/9/2020 11:30 AM, Dmitry Osipenko wrote: >>>> External email: Use caution opening links or attachments >>>> >>>> >>>> 09.01.2020 08:09, Sameer Pujar пишет: >>>>> This patch adds a macro 'regmap_read_poll_timeout_atomic' that works >>>>> similar to 'readx_poll_timeout_atomic' defined in linux/iopoll.h; This >>>>> is atomic version of already available 'regmap_read_poll_timeout' >>>>> macro. >>>>> >>>>> It should be noted that above atomic macro cannot be used by all >>>>> regmaps. >>>>> If the regmap is set up for atomic use (flat or no cache and MMIO) >>>>> then >>>>> only it can use. >>>>> >>>>> Signed-off-by: Sameer Pujar <spujar@nvidia.com> >>>>> --- >>>> Could you please explain what is the targeted use-case here? >>> I was trying to use regmap_read_poll_timeout() to poll for status change >>> of a register. This resulted in "BUG: scheduling while atomic". The >>> callback function, in which I was trying to use the macro, runs in >>> atomic context. Hence new atomic macro is added. I was checking ALSA >>> playback/capture and trigger() callback had to monitor some register >>> status. >>> >>> In general, the new macro can be used in atomic callbacks where regmap >>> interface is used and polling is required. >>> >> You should send a full patchset because it may turn out that the patch >> which makes use of the new feature isn't correct or maybe the new >> feature isn't really needed. >> >> If there was a previous discussion about the need for this change, then >> you should provide a link to that discussion. >> >> Please note that usually changes without a real use-case in kernel are >> not getting picked up or they are getting removed later on if nobody >> makes use of them, so I assume this is a kind of an RFC patch for now. > > OK. I will send this as part of the complete series. Thank you. > Thanks! Please feel free to add me to the CC list, I'll take a look at the patches.
diff --git a/include/linux/regmap.h b/include/linux/regmap.h index dfe493a..f0a092a 100644 --- a/include/linux/regmap.h +++ b/include/linux/regmap.h @@ -145,6 +145,51 @@ struct reg_sequence { }) /** + * regmap_read_poll_timeout_atomic - Poll until a condition is met or a timeout occurs + * + * @map: Regmap to read from + * @addr: Address to poll + * @val: Unsigned integer variable to read the value into + * @cond: Break condition (usually involving @val) + * @delay_us: Time to udelay between reads in us (0 tight-loops). + * Should be less than ~10us since udelay is used + * (see Documentation/timers/timers-howto.rst). + * @timeout_us: Timeout in us, 0 means never timeout + * + * Returns 0 on success and -ETIMEDOUT upon a timeout or the regmap_read + * error return value in case of a error read. In the two former cases, + * the last read value at @addr is stored in @val. + * + * This is modelled after the readx_poll_timeout_atomic macros in linux/iopoll.h. + * + * Note: In general regmap cannot be used in atomic context. If you want to use + * this macro then first setup your regmap for atomic use (flat or no cache + * and MMIO regmap). + */ +#define regmap_read_poll_timeout_atomic(map, addr, val, cond, delay_us, timeout_us) \ +({ \ + u64 __timeout_us = (timeout_us); \ + unsigned long __delay_us = (delay_us); \ + ktime_t __timeout = ktime_add_us(ktime_get(), __timeout_us); \ + int __ret; \ + for (;;) { \ + __ret = regmap_read((map), (addr), &(val)); \ + if (__ret) \ + break; \ + if (cond) \ + break; \ + if ((__timeout_us) && \ + ktime_compare(ktime_get(), __timeout) > 0) { \ + __ret = regmap_read((map), (addr), &(val)); \ + break; \ + } \ + if (__delay_us) \ + udelay(__delay_us); \ + } \ + __ret ?: ((cond) ? 0 : -ETIMEDOUT); \ +}) + +/** * regmap_field_read_poll_timeout - Poll until a condition is met or timeout * * @field: Regmap field to read from
This patch adds a macro 'regmap_read_poll_timeout_atomic' that works similar to 'readx_poll_timeout_atomic' defined in linux/iopoll.h; This is atomic version of already available 'regmap_read_poll_timeout' macro. It should be noted that above atomic macro cannot be used by all regmaps. If the regmap is set up for atomic use (flat or no cache and MMIO) then only it can use. Signed-off-by: Sameer Pujar <spujar@nvidia.com> --- include/linux/regmap.h | 45 +++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 45 insertions(+)