linux-kernelorg-stable/include/linux/iopoll.h

264 lines
10 KiB
C
Raw Normal View History

/* SPDX-License-Identifier: GPL-2.0-only */
/*
* Copyright (c) 2012-2014 The Linux Foundation. All rights reserved.
*/
#ifndef _LINUX_IOPOLL_H
#define _LINUX_IOPOLL_H
#include <linux/kernel.h>
#include <linux/types.h>
#include <linux/ktime.h>
#include <linux/delay.h>
#include <linux/errno.h>
#include <linux/io.h>
/**
iopoll: Generalize read_poll_timeout() into poll_timeout_us() While read_poll_timeout() & co. were originally introduced just for simple I/O usage scenarios they have since been generalized to be useful in more cases. However the interface is very cumbersome to use in the general case. Attempt to make it more flexible by combining the 'op', 'var' and 'args' parameter into just a single 'op' that the caller can fully specify. For example i915 has one case where one might currently have to write something like: ret = read_poll_timeout(drm_dp_dpcd_read_byte, err, err || (status & mask), 0 * 1000, 200 * 1000, false, aux, DP_FEC_STATUS, &status); which is practically illegible, but with the adjusted macro we do: ret = poll_timeout_us(err = drm_dp_dpcd_read_byte(aux, DP_FEC_STATUS, &status), err || (status & mask), 0 * 1000, 200 * 1000, false); which much easier to understand. One could even combine the 'op' and 'cond' parameters into one, but that might make the caller a bit too unwieldly with assignments and checks being done on the same statement. This makes poll_timeout_us() closer to the i915 __wait_for() macro, with the main difference being that __wait_for() uses expenential backoff as opposed to the fixed polling interval used by poll_timeout_us(). Eventually we might be able to switch (at least most of) i915 to use poll_timeout_us(). v2: Fix typos (Jani) Fix delay_us docs for poll_timeout_us_atomic() (Jani) Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com> Cc: Imre Deak <imre.deak@intel.com> Cc: David Laight <david.laight.linux@gmail.com> Cc: Geert Uytterhoeven <geert+renesas@glider.be> Cc: Matt Wagantall <mattw@codeaurora.org> Cc: Dejin Zheng <zhengdejin5@gmail.com> Cc: intel-gfx@lists.freedesktop.org Cc: intel-xe@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Reviewed-by: Jani Nikula <jani.nikula@intel.com> Acked-by: Simona Vetter <simona.vetter@ffwll.ch> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://lore.kernel.org/r/20250826121859.15497-1-ville.syrjala@linux.intel.com Signed-off-by: Jani Nikula <jani.nikula@intel.com>
2025-08-26 12:18:57 +00:00
* poll_timeout_us - Periodically poll and perform an operation until
* a condition is met or a timeout occurs
*
* @op: Operation
* @cond: Break condition
* @sleep_us: Maximum time to sleep between operations in us (0 tight-loops).
* Please read usleep_range() function description for details and
* limitations.
* @timeout_us: Timeout in us, 0 means never timeout
iopoll: Generalize read_poll_timeout() into poll_timeout_us() While read_poll_timeout() & co. were originally introduced just for simple I/O usage scenarios they have since been generalized to be useful in more cases. However the interface is very cumbersome to use in the general case. Attempt to make it more flexible by combining the 'op', 'var' and 'args' parameter into just a single 'op' that the caller can fully specify. For example i915 has one case where one might currently have to write something like: ret = read_poll_timeout(drm_dp_dpcd_read_byte, err, err || (status & mask), 0 * 1000, 200 * 1000, false, aux, DP_FEC_STATUS, &status); which is practically illegible, but with the adjusted macro we do: ret = poll_timeout_us(err = drm_dp_dpcd_read_byte(aux, DP_FEC_STATUS, &status), err || (status & mask), 0 * 1000, 200 * 1000, false); which much easier to understand. One could even combine the 'op' and 'cond' parameters into one, but that might make the caller a bit too unwieldly with assignments and checks being done on the same statement. This makes poll_timeout_us() closer to the i915 __wait_for() macro, with the main difference being that __wait_for() uses expenential backoff as opposed to the fixed polling interval used by poll_timeout_us(). Eventually we might be able to switch (at least most of) i915 to use poll_timeout_us(). v2: Fix typos (Jani) Fix delay_us docs for poll_timeout_us_atomic() (Jani) Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com> Cc: Imre Deak <imre.deak@intel.com> Cc: David Laight <david.laight.linux@gmail.com> Cc: Geert Uytterhoeven <geert+renesas@glider.be> Cc: Matt Wagantall <mattw@codeaurora.org> Cc: Dejin Zheng <zhengdejin5@gmail.com> Cc: intel-gfx@lists.freedesktop.org Cc: intel-xe@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Reviewed-by: Jani Nikula <jani.nikula@intel.com> Acked-by: Simona Vetter <simona.vetter@ffwll.ch> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://lore.kernel.org/r/20250826121859.15497-1-ville.syrjala@linux.intel.com Signed-off-by: Jani Nikula <jani.nikula@intel.com>
2025-08-26 12:18:57 +00:00
* @sleep_before_op: if it is true, sleep @sleep_us before operation.
*
* When available, you'll probably want to use one of the specialized
* macros defined below rather than this macro directly.
*
iopoll: Generalize read_poll_timeout() into poll_timeout_us() While read_poll_timeout() & co. were originally introduced just for simple I/O usage scenarios they have since been generalized to be useful in more cases. However the interface is very cumbersome to use in the general case. Attempt to make it more flexible by combining the 'op', 'var' and 'args' parameter into just a single 'op' that the caller can fully specify. For example i915 has one case where one might currently have to write something like: ret = read_poll_timeout(drm_dp_dpcd_read_byte, err, err || (status & mask), 0 * 1000, 200 * 1000, false, aux, DP_FEC_STATUS, &status); which is practically illegible, but with the adjusted macro we do: ret = poll_timeout_us(err = drm_dp_dpcd_read_byte(aux, DP_FEC_STATUS, &status), err || (status & mask), 0 * 1000, 200 * 1000, false); which much easier to understand. One could even combine the 'op' and 'cond' parameters into one, but that might make the caller a bit too unwieldly with assignments and checks being done on the same statement. This makes poll_timeout_us() closer to the i915 __wait_for() macro, with the main difference being that __wait_for() uses expenential backoff as opposed to the fixed polling interval used by poll_timeout_us(). Eventually we might be able to switch (at least most of) i915 to use poll_timeout_us(). v2: Fix typos (Jani) Fix delay_us docs for poll_timeout_us_atomic() (Jani) Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com> Cc: Imre Deak <imre.deak@intel.com> Cc: David Laight <david.laight.linux@gmail.com> Cc: Geert Uytterhoeven <geert+renesas@glider.be> Cc: Matt Wagantall <mattw@codeaurora.org> Cc: Dejin Zheng <zhengdejin5@gmail.com> Cc: intel-gfx@lists.freedesktop.org Cc: intel-xe@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Reviewed-by: Jani Nikula <jani.nikula@intel.com> Acked-by: Simona Vetter <simona.vetter@ffwll.ch> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://lore.kernel.org/r/20250826121859.15497-1-ville.syrjala@linux.intel.com Signed-off-by: Jani Nikula <jani.nikula@intel.com>
2025-08-26 12:18:57 +00:00
* Returns: 0 on success and -ETIMEDOUT upon a timeout. Must not
* be called from atomic context if sleep_us or timeout_us are used.
*/
iopoll: Generalize read_poll_timeout() into poll_timeout_us() While read_poll_timeout() & co. were originally introduced just for simple I/O usage scenarios they have since been generalized to be useful in more cases. However the interface is very cumbersome to use in the general case. Attempt to make it more flexible by combining the 'op', 'var' and 'args' parameter into just a single 'op' that the caller can fully specify. For example i915 has one case where one might currently have to write something like: ret = read_poll_timeout(drm_dp_dpcd_read_byte, err, err || (status & mask), 0 * 1000, 200 * 1000, false, aux, DP_FEC_STATUS, &status); which is practically illegible, but with the adjusted macro we do: ret = poll_timeout_us(err = drm_dp_dpcd_read_byte(aux, DP_FEC_STATUS, &status), err || (status & mask), 0 * 1000, 200 * 1000, false); which much easier to understand. One could even combine the 'op' and 'cond' parameters into one, but that might make the caller a bit too unwieldly with assignments and checks being done on the same statement. This makes poll_timeout_us() closer to the i915 __wait_for() macro, with the main difference being that __wait_for() uses expenential backoff as opposed to the fixed polling interval used by poll_timeout_us(). Eventually we might be able to switch (at least most of) i915 to use poll_timeout_us(). v2: Fix typos (Jani) Fix delay_us docs for poll_timeout_us_atomic() (Jani) Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com> Cc: Imre Deak <imre.deak@intel.com> Cc: David Laight <david.laight.linux@gmail.com> Cc: Geert Uytterhoeven <geert+renesas@glider.be> Cc: Matt Wagantall <mattw@codeaurora.org> Cc: Dejin Zheng <zhengdejin5@gmail.com> Cc: intel-gfx@lists.freedesktop.org Cc: intel-xe@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Reviewed-by: Jani Nikula <jani.nikula@intel.com> Acked-by: Simona Vetter <simona.vetter@ffwll.ch> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://lore.kernel.org/r/20250826121859.15497-1-ville.syrjala@linux.intel.com Signed-off-by: Jani Nikula <jani.nikula@intel.com>
2025-08-26 12:18:57 +00:00
#define poll_timeout_us(op, cond, sleep_us, timeout_us, sleep_before_op) \
({ \
u64 __timeout_us = (timeout_us); \
unsigned long __sleep_us = (sleep_us); \
ktime_t __timeout = ktime_add_us(ktime_get(), __timeout_us); \
int ___ret; \
might_sleep_if((__sleep_us) != 0); \
iopoll: Generalize read_poll_timeout() into poll_timeout_us() While read_poll_timeout() & co. were originally introduced just for simple I/O usage scenarios they have since been generalized to be useful in more cases. However the interface is very cumbersome to use in the general case. Attempt to make it more flexible by combining the 'op', 'var' and 'args' parameter into just a single 'op' that the caller can fully specify. For example i915 has one case where one might currently have to write something like: ret = read_poll_timeout(drm_dp_dpcd_read_byte, err, err || (status & mask), 0 * 1000, 200 * 1000, false, aux, DP_FEC_STATUS, &status); which is practically illegible, but with the adjusted macro we do: ret = poll_timeout_us(err = drm_dp_dpcd_read_byte(aux, DP_FEC_STATUS, &status), err || (status & mask), 0 * 1000, 200 * 1000, false); which much easier to understand. One could even combine the 'op' and 'cond' parameters into one, but that might make the caller a bit too unwieldly with assignments and checks being done on the same statement. This makes poll_timeout_us() closer to the i915 __wait_for() macro, with the main difference being that __wait_for() uses expenential backoff as opposed to the fixed polling interval used by poll_timeout_us(). Eventually we might be able to switch (at least most of) i915 to use poll_timeout_us(). v2: Fix typos (Jani) Fix delay_us docs for poll_timeout_us_atomic() (Jani) Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com> Cc: Imre Deak <imre.deak@intel.com> Cc: David Laight <david.laight.linux@gmail.com> Cc: Geert Uytterhoeven <geert+renesas@glider.be> Cc: Matt Wagantall <mattw@codeaurora.org> Cc: Dejin Zheng <zhengdejin5@gmail.com> Cc: intel-gfx@lists.freedesktop.org Cc: intel-xe@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Reviewed-by: Jani Nikula <jani.nikula@intel.com> Acked-by: Simona Vetter <simona.vetter@ffwll.ch> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://lore.kernel.org/r/20250826121859.15497-1-ville.syrjala@linux.intel.com Signed-off-by: Jani Nikula <jani.nikula@intel.com>
2025-08-26 12:18:57 +00:00
if ((sleep_before_op) && __sleep_us) \
usleep_range((__sleep_us >> 2) + 1, __sleep_us); \
for (;;) { \
iopoll: Reorder the timeout handling in poll_timeout_us() Currently poll_timeout_us() evaluates 'op' and 'cond' twice within the loop, once at the start, and a second time after the timeout check. While it's probably not a big deal to do it twice almost back to back, it does make the macro a bit messy. Simplify the implementation to evaluate the timeout at the very start, then follow up with 'op'/'cond', and finally check if the timeout did in fact happen or not. For good measure throw in a compiler barrier between the timeout and 'op'/'cond' evaluations to make sure the compiler can't reoder the operations (which could cause false positive timeouts). The similar i915 __wait_for() macro already has the barrier, though there it is between the 'op' and 'cond' evaluations, which seems like it could still allow 'op' and the timeout evaluations to get reordered incorrectly. I suppose the ktime_get() might itself act as a sufficient barrier here, but better safe than sorry I guess. Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com> Cc: Imre Deak <imre.deak@intel.com> Cc: David Laight <david.laight.linux@gmail.com> Cc: Geert Uytterhoeven <geert+renesas@glider.be> Cc: Matt Wagantall <mattw@codeaurora.org> Cc: Dejin Zheng <zhengdejin5@gmail.com> Cc: intel-gfx@lists.freedesktop.org Cc: intel-xe@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Reviewed-by: Jani Nikula <jani.nikula@intel.com> Acked-by: Simona Vetter <simona.vetter@ffwll.ch> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://lore.kernel.org/r/20250826121859.15497-3-ville.syrjala@linux.intel.com Signed-off-by: Jani Nikula <jani.nikula@intel.com>
2025-08-26 12:18:59 +00:00
bool __expired = __timeout_us && \
ktime_compare(ktime_get(), __timeout) > 0; \
/* guarantee 'op' and 'cond' are evaluated after timeout expired */ \
barrier(); \
iopoll: Generalize read_poll_timeout() into poll_timeout_us() While read_poll_timeout() & co. were originally introduced just for simple I/O usage scenarios they have since been generalized to be useful in more cases. However the interface is very cumbersome to use in the general case. Attempt to make it more flexible by combining the 'op', 'var' and 'args' parameter into just a single 'op' that the caller can fully specify. For example i915 has one case where one might currently have to write something like: ret = read_poll_timeout(drm_dp_dpcd_read_byte, err, err || (status & mask), 0 * 1000, 200 * 1000, false, aux, DP_FEC_STATUS, &status); which is practically illegible, but with the adjusted macro we do: ret = poll_timeout_us(err = drm_dp_dpcd_read_byte(aux, DP_FEC_STATUS, &status), err || (status & mask), 0 * 1000, 200 * 1000, false); which much easier to understand. One could even combine the 'op' and 'cond' parameters into one, but that might make the caller a bit too unwieldly with assignments and checks being done on the same statement. This makes poll_timeout_us() closer to the i915 __wait_for() macro, with the main difference being that __wait_for() uses expenential backoff as opposed to the fixed polling interval used by poll_timeout_us(). Eventually we might be able to switch (at least most of) i915 to use poll_timeout_us(). v2: Fix typos (Jani) Fix delay_us docs for poll_timeout_us_atomic() (Jani) Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com> Cc: Imre Deak <imre.deak@intel.com> Cc: David Laight <david.laight.linux@gmail.com> Cc: Geert Uytterhoeven <geert+renesas@glider.be> Cc: Matt Wagantall <mattw@codeaurora.org> Cc: Dejin Zheng <zhengdejin5@gmail.com> Cc: intel-gfx@lists.freedesktop.org Cc: intel-xe@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Reviewed-by: Jani Nikula <jani.nikula@intel.com> Acked-by: Simona Vetter <simona.vetter@ffwll.ch> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://lore.kernel.org/r/20250826121859.15497-1-ville.syrjala@linux.intel.com Signed-off-by: Jani Nikula <jani.nikula@intel.com>
2025-08-26 12:18:57 +00:00
op; \
if (cond) { \
___ret = 0; \
break; \
} \
iopoll: Reorder the timeout handling in poll_timeout_us() Currently poll_timeout_us() evaluates 'op' and 'cond' twice within the loop, once at the start, and a second time after the timeout check. While it's probably not a big deal to do it twice almost back to back, it does make the macro a bit messy. Simplify the implementation to evaluate the timeout at the very start, then follow up with 'op'/'cond', and finally check if the timeout did in fact happen or not. For good measure throw in a compiler barrier between the timeout and 'op'/'cond' evaluations to make sure the compiler can't reoder the operations (which could cause false positive timeouts). The similar i915 __wait_for() macro already has the barrier, though there it is between the 'op' and 'cond' evaluations, which seems like it could still allow 'op' and the timeout evaluations to get reordered incorrectly. I suppose the ktime_get() might itself act as a sufficient barrier here, but better safe than sorry I guess. Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com> Cc: Imre Deak <imre.deak@intel.com> Cc: David Laight <david.laight.linux@gmail.com> Cc: Geert Uytterhoeven <geert+renesas@glider.be> Cc: Matt Wagantall <mattw@codeaurora.org> Cc: Dejin Zheng <zhengdejin5@gmail.com> Cc: intel-gfx@lists.freedesktop.org Cc: intel-xe@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Reviewed-by: Jani Nikula <jani.nikula@intel.com> Acked-by: Simona Vetter <simona.vetter@ffwll.ch> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://lore.kernel.org/r/20250826121859.15497-3-ville.syrjala@linux.intel.com Signed-off-by: Jani Nikula <jani.nikula@intel.com>
2025-08-26 12:18:59 +00:00
if (__expired) { \
___ret = -ETIMEDOUT; \
break; \
} \
if (__sleep_us) \
usleep_range((__sleep_us >> 2) + 1, __sleep_us); \
iopoll: Call cpu_relax() in busy loops It is considered good practice to call cpu_relax() in busy loops, see Documentation/process/volatile-considered-harmful.rst. This can not only lower CPU power consumption or yield to a hyperthreaded twin processor, but also allows an architecture to mitigate hardware issues (e.g. ARM Erratum 754327 for Cortex-A9 prior to r2p0) in the architecture-specific cpu_relax() implementation. In addition, cpu_relax() is also a compiler barrier. It is not immediately obvious that the @op argument "function" will result in an actual function call (e.g. in case of inlining). Where a function call is a C sequence point, this is lost on inlining. Therefore, with agressive enough optimization it might be possible for the compiler to hoist the: (val) = op(args); "load" out of the loop because it doesn't see the value changing. The addition of cpu_relax() would inhibit this. As the iopoll helpers lack calls to cpu_relax(), people are sometimes reluctant to use them, and may fall back to open-coded polling loops (including cpu_relax() calls) instead. Fix this by adding calls to cpu_relax() to the iopoll helpers: - For the non-atomic case, it is sufficient to call cpu_relax() in case of a zero sleep-between-reads value, as a call to usleep_range() is a safe barrier otherwise. However, it doesn't hurt to add the call regardless, for simplicity, and for similarity with the atomic case below. - For the atomic case, cpu_relax() must be called regardless of the sleep-between-reads value, as there is no guarantee all architecture-specific implementations of udelay() handle this. Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Arnd Bergmann <arnd@arndb.de> Reviewed-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org> Link: https://lore.kernel.org/r/45c87bec3397fdd704376807f0eec5cc71be440f.1685692810.git.geert+renesas@glider.be
2023-06-02 08:50:36 +00:00
cpu_relax(); \
} \
___ret; \
})
/**
iopoll: Generalize read_poll_timeout() into poll_timeout_us() While read_poll_timeout() & co. were originally introduced just for simple I/O usage scenarios they have since been generalized to be useful in more cases. However the interface is very cumbersome to use in the general case. Attempt to make it more flexible by combining the 'op', 'var' and 'args' parameter into just a single 'op' that the caller can fully specify. For example i915 has one case where one might currently have to write something like: ret = read_poll_timeout(drm_dp_dpcd_read_byte, err, err || (status & mask), 0 * 1000, 200 * 1000, false, aux, DP_FEC_STATUS, &status); which is practically illegible, but with the adjusted macro we do: ret = poll_timeout_us(err = drm_dp_dpcd_read_byte(aux, DP_FEC_STATUS, &status), err || (status & mask), 0 * 1000, 200 * 1000, false); which much easier to understand. One could even combine the 'op' and 'cond' parameters into one, but that might make the caller a bit too unwieldly with assignments and checks being done on the same statement. This makes poll_timeout_us() closer to the i915 __wait_for() macro, with the main difference being that __wait_for() uses expenential backoff as opposed to the fixed polling interval used by poll_timeout_us(). Eventually we might be able to switch (at least most of) i915 to use poll_timeout_us(). v2: Fix typos (Jani) Fix delay_us docs for poll_timeout_us_atomic() (Jani) Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com> Cc: Imre Deak <imre.deak@intel.com> Cc: David Laight <david.laight.linux@gmail.com> Cc: Geert Uytterhoeven <geert+renesas@glider.be> Cc: Matt Wagantall <mattw@codeaurora.org> Cc: Dejin Zheng <zhengdejin5@gmail.com> Cc: intel-gfx@lists.freedesktop.org Cc: intel-xe@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Reviewed-by: Jani Nikula <jani.nikula@intel.com> Acked-by: Simona Vetter <simona.vetter@ffwll.ch> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://lore.kernel.org/r/20250826121859.15497-1-ville.syrjala@linux.intel.com Signed-off-by: Jani Nikula <jani.nikula@intel.com>
2025-08-26 12:18:57 +00:00
* poll_timeout_us_atomic - Periodically poll and perform an operation until
* a condition is met or a timeout occurs
*
* @op: Operation
* @cond: Break condition
* @delay_us: Time to udelay between operations in us (0 tight-loops).
* Please read udelay() function description for details and
* limitations.
* @timeout_us: Timeout in us, 0 means never timeout
iopoll: Generalize read_poll_timeout() into poll_timeout_us() While read_poll_timeout() & co. were originally introduced just for simple I/O usage scenarios they have since been generalized to be useful in more cases. However the interface is very cumbersome to use in the general case. Attempt to make it more flexible by combining the 'op', 'var' and 'args' parameter into just a single 'op' that the caller can fully specify. For example i915 has one case where one might currently have to write something like: ret = read_poll_timeout(drm_dp_dpcd_read_byte, err, err || (status & mask), 0 * 1000, 200 * 1000, false, aux, DP_FEC_STATUS, &status); which is practically illegible, but with the adjusted macro we do: ret = poll_timeout_us(err = drm_dp_dpcd_read_byte(aux, DP_FEC_STATUS, &status), err || (status & mask), 0 * 1000, 200 * 1000, false); which much easier to understand. One could even combine the 'op' and 'cond' parameters into one, but that might make the caller a bit too unwieldly with assignments and checks being done on the same statement. This makes poll_timeout_us() closer to the i915 __wait_for() macro, with the main difference being that __wait_for() uses expenential backoff as opposed to the fixed polling interval used by poll_timeout_us(). Eventually we might be able to switch (at least most of) i915 to use poll_timeout_us(). v2: Fix typos (Jani) Fix delay_us docs for poll_timeout_us_atomic() (Jani) Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com> Cc: Imre Deak <imre.deak@intel.com> Cc: David Laight <david.laight.linux@gmail.com> Cc: Geert Uytterhoeven <geert+renesas@glider.be> Cc: Matt Wagantall <mattw@codeaurora.org> Cc: Dejin Zheng <zhengdejin5@gmail.com> Cc: intel-gfx@lists.freedesktop.org Cc: intel-xe@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Reviewed-by: Jani Nikula <jani.nikula@intel.com> Acked-by: Simona Vetter <simona.vetter@ffwll.ch> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://lore.kernel.org/r/20250826121859.15497-1-ville.syrjala@linux.intel.com Signed-off-by: Jani Nikula <jani.nikula@intel.com>
2025-08-26 12:18:57 +00:00
* @delay_before_op: if it is true, delay @delay_us before operation.
*
iopoll: Do not use timekeeping in read_poll_timeout_atomic() read_poll_timeout_atomic() uses ktime_get() to implement the timeout feature, just like its non-atomic counterpart. However, there are several issues with this, due to its use in atomic contexts: 1. When called in the s2ram path (as typically done by clock or PM domain drivers), timekeeping may be suspended, triggering the WARN_ON(timekeeping_suspended) in ktime_get(): WARNING: CPU: 0 PID: 654 at kernel/time/timekeeping.c:843 ktime_get+0x28/0x78 Calling ktime_get_mono_fast_ns() instead of ktime_get() would get rid of that warning. However, that would break timeout handling, as (at least on systems with an ARM architectured timer), the time returned by ktime_get_mono_fast_ns() does not advance while timekeeping is suspended. Interestingly, (on the same ARM systems) the time returned by ktime_get() does advance while timekeeping is suspended, despite the warning. 2. Depending on the actual clock source, and especially before a high-resolution clocksource (e.g. the ARM architectured timer) becomes available, time may not advance in atomic contexts, thus breaking timeout handling. Fix this by abandoning the idea that one can rely on timekeeping to implement timeout handling in all atomic contexts, and switch from a global time-based to a locally-estimated timeout handling. In most (all?) cases the timeout condition is exceptional and an error condition, hence any additional delays due to underestimating wall clock time are irrelevant. Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be> Acked-by: Arnd Bergmann <arnd@arndb.de> Reviewed-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org> Link: https://lore.kernel.org/r/3d2a2f4e553489392d871108797c3be08f88300b.1685692810.git.geert+renesas@glider.be
2023-06-02 08:50:37 +00:00
* This macro does not rely on timekeeping. Hence it is safe to call even when
* timekeeping is suspended, at the expense of an underestimation of wall clock
* time, which is rather minimal with a non-zero delay_us.
*
* When available, you'll probably want to use one of the specialized
* macros defined below rather than this macro directly.
*
iopoll: Generalize read_poll_timeout() into poll_timeout_us() While read_poll_timeout() & co. were originally introduced just for simple I/O usage scenarios they have since been generalized to be useful in more cases. However the interface is very cumbersome to use in the general case. Attempt to make it more flexible by combining the 'op', 'var' and 'args' parameter into just a single 'op' that the caller can fully specify. For example i915 has one case where one might currently have to write something like: ret = read_poll_timeout(drm_dp_dpcd_read_byte, err, err || (status & mask), 0 * 1000, 200 * 1000, false, aux, DP_FEC_STATUS, &status); which is practically illegible, but with the adjusted macro we do: ret = poll_timeout_us(err = drm_dp_dpcd_read_byte(aux, DP_FEC_STATUS, &status), err || (status & mask), 0 * 1000, 200 * 1000, false); which much easier to understand. One could even combine the 'op' and 'cond' parameters into one, but that might make the caller a bit too unwieldly with assignments and checks being done on the same statement. This makes poll_timeout_us() closer to the i915 __wait_for() macro, with the main difference being that __wait_for() uses expenential backoff as opposed to the fixed polling interval used by poll_timeout_us(). Eventually we might be able to switch (at least most of) i915 to use poll_timeout_us(). v2: Fix typos (Jani) Fix delay_us docs for poll_timeout_us_atomic() (Jani) Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com> Cc: Imre Deak <imre.deak@intel.com> Cc: David Laight <david.laight.linux@gmail.com> Cc: Geert Uytterhoeven <geert+renesas@glider.be> Cc: Matt Wagantall <mattw@codeaurora.org> Cc: Dejin Zheng <zhengdejin5@gmail.com> Cc: intel-gfx@lists.freedesktop.org Cc: intel-xe@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Reviewed-by: Jani Nikula <jani.nikula@intel.com> Acked-by: Simona Vetter <simona.vetter@ffwll.ch> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://lore.kernel.org/r/20250826121859.15497-1-ville.syrjala@linux.intel.com Signed-off-by: Jani Nikula <jani.nikula@intel.com>
2025-08-26 12:18:57 +00:00
* Returns: 0 on success and -ETIMEDOUT upon a timeout.
*/
iopoll: Generalize read_poll_timeout() into poll_timeout_us() While read_poll_timeout() & co. were originally introduced just for simple I/O usage scenarios they have since been generalized to be useful in more cases. However the interface is very cumbersome to use in the general case. Attempt to make it more flexible by combining the 'op', 'var' and 'args' parameter into just a single 'op' that the caller can fully specify. For example i915 has one case where one might currently have to write something like: ret = read_poll_timeout(drm_dp_dpcd_read_byte, err, err || (status & mask), 0 * 1000, 200 * 1000, false, aux, DP_FEC_STATUS, &status); which is practically illegible, but with the adjusted macro we do: ret = poll_timeout_us(err = drm_dp_dpcd_read_byte(aux, DP_FEC_STATUS, &status), err || (status & mask), 0 * 1000, 200 * 1000, false); which much easier to understand. One could even combine the 'op' and 'cond' parameters into one, but that might make the caller a bit too unwieldly with assignments and checks being done on the same statement. This makes poll_timeout_us() closer to the i915 __wait_for() macro, with the main difference being that __wait_for() uses expenential backoff as opposed to the fixed polling interval used by poll_timeout_us(). Eventually we might be able to switch (at least most of) i915 to use poll_timeout_us(). v2: Fix typos (Jani) Fix delay_us docs for poll_timeout_us_atomic() (Jani) Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com> Cc: Imre Deak <imre.deak@intel.com> Cc: David Laight <david.laight.linux@gmail.com> Cc: Geert Uytterhoeven <geert+renesas@glider.be> Cc: Matt Wagantall <mattw@codeaurora.org> Cc: Dejin Zheng <zhengdejin5@gmail.com> Cc: intel-gfx@lists.freedesktop.org Cc: intel-xe@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Reviewed-by: Jani Nikula <jani.nikula@intel.com> Acked-by: Simona Vetter <simona.vetter@ffwll.ch> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://lore.kernel.org/r/20250826121859.15497-1-ville.syrjala@linux.intel.com Signed-off-by: Jani Nikula <jani.nikula@intel.com>
2025-08-26 12:18:57 +00:00
#define poll_timeout_us_atomic(op, cond, delay_us, timeout_us, \
delay_before_op) \
({ \
u64 __timeout_us = (timeout_us); \
iopoll: Do not use timekeeping in read_poll_timeout_atomic() read_poll_timeout_atomic() uses ktime_get() to implement the timeout feature, just like its non-atomic counterpart. However, there are several issues with this, due to its use in atomic contexts: 1. When called in the s2ram path (as typically done by clock or PM domain drivers), timekeeping may be suspended, triggering the WARN_ON(timekeeping_suspended) in ktime_get(): WARNING: CPU: 0 PID: 654 at kernel/time/timekeeping.c:843 ktime_get+0x28/0x78 Calling ktime_get_mono_fast_ns() instead of ktime_get() would get rid of that warning. However, that would break timeout handling, as (at least on systems with an ARM architectured timer), the time returned by ktime_get_mono_fast_ns() does not advance while timekeeping is suspended. Interestingly, (on the same ARM systems) the time returned by ktime_get() does advance while timekeeping is suspended, despite the warning. 2. Depending on the actual clock source, and especially before a high-resolution clocksource (e.g. the ARM architectured timer) becomes available, time may not advance in atomic contexts, thus breaking timeout handling. Fix this by abandoning the idea that one can rely on timekeeping to implement timeout handling in all atomic contexts, and switch from a global time-based to a locally-estimated timeout handling. In most (all?) cases the timeout condition is exceptional and an error condition, hence any additional delays due to underestimating wall clock time are irrelevant. Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be> Acked-by: Arnd Bergmann <arnd@arndb.de> Reviewed-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org> Link: https://lore.kernel.org/r/3d2a2f4e553489392d871108797c3be08f88300b.1685692810.git.geert+renesas@glider.be
2023-06-02 08:50:37 +00:00
s64 __left_ns = __timeout_us * NSEC_PER_USEC; \
unsigned long __delay_us = (delay_us); \
iopoll: Do not use timekeeping in read_poll_timeout_atomic() read_poll_timeout_atomic() uses ktime_get() to implement the timeout feature, just like its non-atomic counterpart. However, there are several issues with this, due to its use in atomic contexts: 1. When called in the s2ram path (as typically done by clock or PM domain drivers), timekeeping may be suspended, triggering the WARN_ON(timekeeping_suspended) in ktime_get(): WARNING: CPU: 0 PID: 654 at kernel/time/timekeeping.c:843 ktime_get+0x28/0x78 Calling ktime_get_mono_fast_ns() instead of ktime_get() would get rid of that warning. However, that would break timeout handling, as (at least on systems with an ARM architectured timer), the time returned by ktime_get_mono_fast_ns() does not advance while timekeeping is suspended. Interestingly, (on the same ARM systems) the time returned by ktime_get() does advance while timekeeping is suspended, despite the warning. 2. Depending on the actual clock source, and especially before a high-resolution clocksource (e.g. the ARM architectured timer) becomes available, time may not advance in atomic contexts, thus breaking timeout handling. Fix this by abandoning the idea that one can rely on timekeeping to implement timeout handling in all atomic contexts, and switch from a global time-based to a locally-estimated timeout handling. In most (all?) cases the timeout condition is exceptional and an error condition, hence any additional delays due to underestimating wall clock time are irrelevant. Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be> Acked-by: Arnd Bergmann <arnd@arndb.de> Reviewed-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org> Link: https://lore.kernel.org/r/3d2a2f4e553489392d871108797c3be08f88300b.1685692810.git.geert+renesas@glider.be
2023-06-02 08:50:37 +00:00
u64 __delay_ns = __delay_us * NSEC_PER_USEC; \
int ___ret; \
iopoll: Generalize read_poll_timeout() into poll_timeout_us() While read_poll_timeout() & co. were originally introduced just for simple I/O usage scenarios they have since been generalized to be useful in more cases. However the interface is very cumbersome to use in the general case. Attempt to make it more flexible by combining the 'op', 'var' and 'args' parameter into just a single 'op' that the caller can fully specify. For example i915 has one case where one might currently have to write something like: ret = read_poll_timeout(drm_dp_dpcd_read_byte, err, err || (status & mask), 0 * 1000, 200 * 1000, false, aux, DP_FEC_STATUS, &status); which is practically illegible, but with the adjusted macro we do: ret = poll_timeout_us(err = drm_dp_dpcd_read_byte(aux, DP_FEC_STATUS, &status), err || (status & mask), 0 * 1000, 200 * 1000, false); which much easier to understand. One could even combine the 'op' and 'cond' parameters into one, but that might make the caller a bit too unwieldly with assignments and checks being done on the same statement. This makes poll_timeout_us() closer to the i915 __wait_for() macro, with the main difference being that __wait_for() uses expenential backoff as opposed to the fixed polling interval used by poll_timeout_us(). Eventually we might be able to switch (at least most of) i915 to use poll_timeout_us(). v2: Fix typos (Jani) Fix delay_us docs for poll_timeout_us_atomic() (Jani) Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com> Cc: Imre Deak <imre.deak@intel.com> Cc: David Laight <david.laight.linux@gmail.com> Cc: Geert Uytterhoeven <geert+renesas@glider.be> Cc: Matt Wagantall <mattw@codeaurora.org> Cc: Dejin Zheng <zhengdejin5@gmail.com> Cc: intel-gfx@lists.freedesktop.org Cc: intel-xe@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Reviewed-by: Jani Nikula <jani.nikula@intel.com> Acked-by: Simona Vetter <simona.vetter@ffwll.ch> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://lore.kernel.org/r/20250826121859.15497-1-ville.syrjala@linux.intel.com Signed-off-by: Jani Nikula <jani.nikula@intel.com>
2025-08-26 12:18:57 +00:00
if ((delay_before_op) && __delay_us) { \
udelay(__delay_us); \
iopoll: Do not use timekeeping in read_poll_timeout_atomic() read_poll_timeout_atomic() uses ktime_get() to implement the timeout feature, just like its non-atomic counterpart. However, there are several issues with this, due to its use in atomic contexts: 1. When called in the s2ram path (as typically done by clock or PM domain drivers), timekeeping may be suspended, triggering the WARN_ON(timekeeping_suspended) in ktime_get(): WARNING: CPU: 0 PID: 654 at kernel/time/timekeeping.c:843 ktime_get+0x28/0x78 Calling ktime_get_mono_fast_ns() instead of ktime_get() would get rid of that warning. However, that would break timeout handling, as (at least on systems with an ARM architectured timer), the time returned by ktime_get_mono_fast_ns() does not advance while timekeeping is suspended. Interestingly, (on the same ARM systems) the time returned by ktime_get() does advance while timekeeping is suspended, despite the warning. 2. Depending on the actual clock source, and especially before a high-resolution clocksource (e.g. the ARM architectured timer) becomes available, time may not advance in atomic contexts, thus breaking timeout handling. Fix this by abandoning the idea that one can rely on timekeeping to implement timeout handling in all atomic contexts, and switch from a global time-based to a locally-estimated timeout handling. In most (all?) cases the timeout condition is exceptional and an error condition, hence any additional delays due to underestimating wall clock time are irrelevant. Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be> Acked-by: Arnd Bergmann <arnd@arndb.de> Reviewed-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org> Link: https://lore.kernel.org/r/3d2a2f4e553489392d871108797c3be08f88300b.1685692810.git.geert+renesas@glider.be
2023-06-02 08:50:37 +00:00
if (__timeout_us) \
__left_ns -= __delay_ns; \
} \
for (;;) { \
iopoll: Reorder the timeout handling in poll_timeout_us() Currently poll_timeout_us() evaluates 'op' and 'cond' twice within the loop, once at the start, and a second time after the timeout check. While it's probably not a big deal to do it twice almost back to back, it does make the macro a bit messy. Simplify the implementation to evaluate the timeout at the very start, then follow up with 'op'/'cond', and finally check if the timeout did in fact happen or not. For good measure throw in a compiler barrier between the timeout and 'op'/'cond' evaluations to make sure the compiler can't reoder the operations (which could cause false positive timeouts). The similar i915 __wait_for() macro already has the barrier, though there it is between the 'op' and 'cond' evaluations, which seems like it could still allow 'op' and the timeout evaluations to get reordered incorrectly. I suppose the ktime_get() might itself act as a sufficient barrier here, but better safe than sorry I guess. Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com> Cc: Imre Deak <imre.deak@intel.com> Cc: David Laight <david.laight.linux@gmail.com> Cc: Geert Uytterhoeven <geert+renesas@glider.be> Cc: Matt Wagantall <mattw@codeaurora.org> Cc: Dejin Zheng <zhengdejin5@gmail.com> Cc: intel-gfx@lists.freedesktop.org Cc: intel-xe@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Reviewed-by: Jani Nikula <jani.nikula@intel.com> Acked-by: Simona Vetter <simona.vetter@ffwll.ch> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://lore.kernel.org/r/20250826121859.15497-3-ville.syrjala@linux.intel.com Signed-off-by: Jani Nikula <jani.nikula@intel.com>
2025-08-26 12:18:59 +00:00
bool __expired = __timeout_us && __left_ns < 0; \
/* guarantee 'op' and 'cond' are evaluated after timeout expired */ \
barrier(); \
iopoll: Generalize read_poll_timeout() into poll_timeout_us() While read_poll_timeout() & co. were originally introduced just for simple I/O usage scenarios they have since been generalized to be useful in more cases. However the interface is very cumbersome to use in the general case. Attempt to make it more flexible by combining the 'op', 'var' and 'args' parameter into just a single 'op' that the caller can fully specify. For example i915 has one case where one might currently have to write something like: ret = read_poll_timeout(drm_dp_dpcd_read_byte, err, err || (status & mask), 0 * 1000, 200 * 1000, false, aux, DP_FEC_STATUS, &status); which is practically illegible, but with the adjusted macro we do: ret = poll_timeout_us(err = drm_dp_dpcd_read_byte(aux, DP_FEC_STATUS, &status), err || (status & mask), 0 * 1000, 200 * 1000, false); which much easier to understand. One could even combine the 'op' and 'cond' parameters into one, but that might make the caller a bit too unwieldly with assignments and checks being done on the same statement. This makes poll_timeout_us() closer to the i915 __wait_for() macro, with the main difference being that __wait_for() uses expenential backoff as opposed to the fixed polling interval used by poll_timeout_us(). Eventually we might be able to switch (at least most of) i915 to use poll_timeout_us(). v2: Fix typos (Jani) Fix delay_us docs for poll_timeout_us_atomic() (Jani) Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com> Cc: Imre Deak <imre.deak@intel.com> Cc: David Laight <david.laight.linux@gmail.com> Cc: Geert Uytterhoeven <geert+renesas@glider.be> Cc: Matt Wagantall <mattw@codeaurora.org> Cc: Dejin Zheng <zhengdejin5@gmail.com> Cc: intel-gfx@lists.freedesktop.org Cc: intel-xe@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Reviewed-by: Jani Nikula <jani.nikula@intel.com> Acked-by: Simona Vetter <simona.vetter@ffwll.ch> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://lore.kernel.org/r/20250826121859.15497-1-ville.syrjala@linux.intel.com Signed-off-by: Jani Nikula <jani.nikula@intel.com>
2025-08-26 12:18:57 +00:00
op; \
if (cond) { \
___ret = 0; \
break; \
} \
iopoll: Reorder the timeout handling in poll_timeout_us() Currently poll_timeout_us() evaluates 'op' and 'cond' twice within the loop, once at the start, and a second time after the timeout check. While it's probably not a big deal to do it twice almost back to back, it does make the macro a bit messy. Simplify the implementation to evaluate the timeout at the very start, then follow up with 'op'/'cond', and finally check if the timeout did in fact happen or not. For good measure throw in a compiler barrier between the timeout and 'op'/'cond' evaluations to make sure the compiler can't reoder the operations (which could cause false positive timeouts). The similar i915 __wait_for() macro already has the barrier, though there it is between the 'op' and 'cond' evaluations, which seems like it could still allow 'op' and the timeout evaluations to get reordered incorrectly. I suppose the ktime_get() might itself act as a sufficient barrier here, but better safe than sorry I guess. Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com> Cc: Imre Deak <imre.deak@intel.com> Cc: David Laight <david.laight.linux@gmail.com> Cc: Geert Uytterhoeven <geert+renesas@glider.be> Cc: Matt Wagantall <mattw@codeaurora.org> Cc: Dejin Zheng <zhengdejin5@gmail.com> Cc: intel-gfx@lists.freedesktop.org Cc: intel-xe@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Reviewed-by: Jani Nikula <jani.nikula@intel.com> Acked-by: Simona Vetter <simona.vetter@ffwll.ch> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://lore.kernel.org/r/20250826121859.15497-3-ville.syrjala@linux.intel.com Signed-off-by: Jani Nikula <jani.nikula@intel.com>
2025-08-26 12:18:59 +00:00
if (__expired) { \
___ret = -ETIMEDOUT; \
break; \
} \
iopoll: Do not use timekeeping in read_poll_timeout_atomic() read_poll_timeout_atomic() uses ktime_get() to implement the timeout feature, just like its non-atomic counterpart. However, there are several issues with this, due to its use in atomic contexts: 1. When called in the s2ram path (as typically done by clock or PM domain drivers), timekeeping may be suspended, triggering the WARN_ON(timekeeping_suspended) in ktime_get(): WARNING: CPU: 0 PID: 654 at kernel/time/timekeeping.c:843 ktime_get+0x28/0x78 Calling ktime_get_mono_fast_ns() instead of ktime_get() would get rid of that warning. However, that would break timeout handling, as (at least on systems with an ARM architectured timer), the time returned by ktime_get_mono_fast_ns() does not advance while timekeeping is suspended. Interestingly, (on the same ARM systems) the time returned by ktime_get() does advance while timekeeping is suspended, despite the warning. 2. Depending on the actual clock source, and especially before a high-resolution clocksource (e.g. the ARM architectured timer) becomes available, time may not advance in atomic contexts, thus breaking timeout handling. Fix this by abandoning the idea that one can rely on timekeeping to implement timeout handling in all atomic contexts, and switch from a global time-based to a locally-estimated timeout handling. In most (all?) cases the timeout condition is exceptional and an error condition, hence any additional delays due to underestimating wall clock time are irrelevant. Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be> Acked-by: Arnd Bergmann <arnd@arndb.de> Reviewed-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org> Link: https://lore.kernel.org/r/3d2a2f4e553489392d871108797c3be08f88300b.1685692810.git.geert+renesas@glider.be
2023-06-02 08:50:37 +00:00
if (__delay_us) { \
udelay(__delay_us); \
iopoll: Do not use timekeeping in read_poll_timeout_atomic() read_poll_timeout_atomic() uses ktime_get() to implement the timeout feature, just like its non-atomic counterpart. However, there are several issues with this, due to its use in atomic contexts: 1. When called in the s2ram path (as typically done by clock or PM domain drivers), timekeeping may be suspended, triggering the WARN_ON(timekeeping_suspended) in ktime_get(): WARNING: CPU: 0 PID: 654 at kernel/time/timekeeping.c:843 ktime_get+0x28/0x78 Calling ktime_get_mono_fast_ns() instead of ktime_get() would get rid of that warning. However, that would break timeout handling, as (at least on systems with an ARM architectured timer), the time returned by ktime_get_mono_fast_ns() does not advance while timekeeping is suspended. Interestingly, (on the same ARM systems) the time returned by ktime_get() does advance while timekeeping is suspended, despite the warning. 2. Depending on the actual clock source, and especially before a high-resolution clocksource (e.g. the ARM architectured timer) becomes available, time may not advance in atomic contexts, thus breaking timeout handling. Fix this by abandoning the idea that one can rely on timekeeping to implement timeout handling in all atomic contexts, and switch from a global time-based to a locally-estimated timeout handling. In most (all?) cases the timeout condition is exceptional and an error condition, hence any additional delays due to underestimating wall clock time are irrelevant. Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be> Acked-by: Arnd Bergmann <arnd@arndb.de> Reviewed-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org> Link: https://lore.kernel.org/r/3d2a2f4e553489392d871108797c3be08f88300b.1685692810.git.geert+renesas@glider.be
2023-06-02 08:50:37 +00:00
if (__timeout_us) \
__left_ns -= __delay_ns; \
} \
iopoll: Call cpu_relax() in busy loops It is considered good practice to call cpu_relax() in busy loops, see Documentation/process/volatile-considered-harmful.rst. This can not only lower CPU power consumption or yield to a hyperthreaded twin processor, but also allows an architecture to mitigate hardware issues (e.g. ARM Erratum 754327 for Cortex-A9 prior to r2p0) in the architecture-specific cpu_relax() implementation. In addition, cpu_relax() is also a compiler barrier. It is not immediately obvious that the @op argument "function" will result in an actual function call (e.g. in case of inlining). Where a function call is a C sequence point, this is lost on inlining. Therefore, with agressive enough optimization it might be possible for the compiler to hoist the: (val) = op(args); "load" out of the loop because it doesn't see the value changing. The addition of cpu_relax() would inhibit this. As the iopoll helpers lack calls to cpu_relax(), people are sometimes reluctant to use them, and may fall back to open-coded polling loops (including cpu_relax() calls) instead. Fix this by adding calls to cpu_relax() to the iopoll helpers: - For the non-atomic case, it is sufficient to call cpu_relax() in case of a zero sleep-between-reads value, as a call to usleep_range() is a safe barrier otherwise. However, it doesn't hurt to add the call regardless, for simplicity, and for similarity with the atomic case below. - For the atomic case, cpu_relax() must be called regardless of the sleep-between-reads value, as there is no guarantee all architecture-specific implementations of udelay() handle this. Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Arnd Bergmann <arnd@arndb.de> Reviewed-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org> Link: https://lore.kernel.org/r/45c87bec3397fdd704376807f0eec5cc71be440f.1685692810.git.geert+renesas@glider.be
2023-06-02 08:50:36 +00:00
cpu_relax(); \
iopoll: Do not use timekeeping in read_poll_timeout_atomic() read_poll_timeout_atomic() uses ktime_get() to implement the timeout feature, just like its non-atomic counterpart. However, there are several issues with this, due to its use in atomic contexts: 1. When called in the s2ram path (as typically done by clock or PM domain drivers), timekeeping may be suspended, triggering the WARN_ON(timekeeping_suspended) in ktime_get(): WARNING: CPU: 0 PID: 654 at kernel/time/timekeeping.c:843 ktime_get+0x28/0x78 Calling ktime_get_mono_fast_ns() instead of ktime_get() would get rid of that warning. However, that would break timeout handling, as (at least on systems with an ARM architectured timer), the time returned by ktime_get_mono_fast_ns() does not advance while timekeeping is suspended. Interestingly, (on the same ARM systems) the time returned by ktime_get() does advance while timekeeping is suspended, despite the warning. 2. Depending on the actual clock source, and especially before a high-resolution clocksource (e.g. the ARM architectured timer) becomes available, time may not advance in atomic contexts, thus breaking timeout handling. Fix this by abandoning the idea that one can rely on timekeeping to implement timeout handling in all atomic contexts, and switch from a global time-based to a locally-estimated timeout handling. In most (all?) cases the timeout condition is exceptional and an error condition, hence any additional delays due to underestimating wall clock time are irrelevant. Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be> Acked-by: Arnd Bergmann <arnd@arndb.de> Reviewed-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org> Link: https://lore.kernel.org/r/3d2a2f4e553489392d871108797c3be08f88300b.1685692810.git.geert+renesas@glider.be
2023-06-02 08:50:37 +00:00
if (__timeout_us) \
__left_ns--; \
} \
___ret; \
})
iopoll: Generalize read_poll_timeout() into poll_timeout_us() While read_poll_timeout() & co. were originally introduced just for simple I/O usage scenarios they have since been generalized to be useful in more cases. However the interface is very cumbersome to use in the general case. Attempt to make it more flexible by combining the 'op', 'var' and 'args' parameter into just a single 'op' that the caller can fully specify. For example i915 has one case where one might currently have to write something like: ret = read_poll_timeout(drm_dp_dpcd_read_byte, err, err || (status & mask), 0 * 1000, 200 * 1000, false, aux, DP_FEC_STATUS, &status); which is practically illegible, but with the adjusted macro we do: ret = poll_timeout_us(err = drm_dp_dpcd_read_byte(aux, DP_FEC_STATUS, &status), err || (status & mask), 0 * 1000, 200 * 1000, false); which much easier to understand. One could even combine the 'op' and 'cond' parameters into one, but that might make the caller a bit too unwieldly with assignments and checks being done on the same statement. This makes poll_timeout_us() closer to the i915 __wait_for() macro, with the main difference being that __wait_for() uses expenential backoff as opposed to the fixed polling interval used by poll_timeout_us(). Eventually we might be able to switch (at least most of) i915 to use poll_timeout_us(). v2: Fix typos (Jani) Fix delay_us docs for poll_timeout_us_atomic() (Jani) Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com> Cc: Imre Deak <imre.deak@intel.com> Cc: David Laight <david.laight.linux@gmail.com> Cc: Geert Uytterhoeven <geert+renesas@glider.be> Cc: Matt Wagantall <mattw@codeaurora.org> Cc: Dejin Zheng <zhengdejin5@gmail.com> Cc: intel-gfx@lists.freedesktop.org Cc: intel-xe@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Reviewed-by: Jani Nikula <jani.nikula@intel.com> Acked-by: Simona Vetter <simona.vetter@ffwll.ch> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://lore.kernel.org/r/20250826121859.15497-1-ville.syrjala@linux.intel.com Signed-off-by: Jani Nikula <jani.nikula@intel.com>
2025-08-26 12:18:57 +00:00
/**
* read_poll_timeout - Periodically poll an address until a condition is
* met or a timeout occurs
* @op: accessor function (takes @args as its arguments)
* @val: Variable to read the value into
* @cond: Break condition (usually involving @val)
* @sleep_us: Maximum time to sleep between reads in us (0 tight-loops). Please
* read usleep_range() function description for details and
* limitations.
* @timeout_us: Timeout in us, 0 means never timeout
* @sleep_before_read: if it is true, sleep @sleep_us before read.
* @args: arguments for @op poll
*
* When available, you'll probably want to use one of the specialized
* macros defined below rather than this macro directly.
*
* Returns: 0 on success and -ETIMEDOUT upon a timeout. In either
* case, the last read value at @args is stored in @val. Must not
* be called from atomic context if sleep_us or timeout_us are used.
*/
#define read_poll_timeout(op, val, cond, sleep_us, timeout_us, \
sleep_before_read, args...) \
poll_timeout_us((val) = op(args), cond, sleep_us, timeout_us, sleep_before_read)
/**
* read_poll_timeout_atomic - Periodically poll an address until a condition is
* met or a timeout occurs
* @op: accessor function (takes @args as its arguments)
* @val: Variable to read the value into
* @cond: Break condition (usually involving @val)
* @delay_us: Time to udelay between reads in us (0 tight-loops). Please
* read udelay() function description for details and
* limitations.
* @timeout_us: Timeout in us, 0 means never timeout
* @delay_before_read: if it is true, delay @delay_us before read.
* @args: arguments for @op poll
*
* This macro does not rely on timekeeping. Hence it is safe to call even when
* timekeeping is suspended, at the expense of an underestimation of wall clock
* time, which is rather minimal with a non-zero delay_us.
*
* When available, you'll probably want to use one of the specialized
* macros defined below rather than this macro directly.
*
* Returns: 0 on success and -ETIMEDOUT upon a timeout. In either
* case, the last read value at @args is stored in @val.
*/
#define read_poll_timeout_atomic(op, val, cond, sleep_us, timeout_us, \
sleep_before_read, args...) \
poll_timeout_us_atomic((val) = op(args), cond, sleep_us, timeout_us, sleep_before_read)
/**
* readx_poll_timeout - Periodically poll an address until a condition is met or a timeout occurs
* @op: accessor function (takes @addr as its only argument)
* @addr: Address to poll
* @val: Variable to read the value into
* @cond: Break condition (usually involving @val)
* @sleep_us: Maximum time to sleep between reads in us (0 tight-loops). Please
* read usleep_range() function description for details and
* limitations.
* @timeout_us: Timeout in us, 0 means never timeout
*
* When available, you'll probably want to use one of the specialized
* macros defined below rather than this macro directly.
*
* Returns: 0 on success and -ETIMEDOUT upon a timeout. In either
* case, the last read value at @addr is stored in @val. Must not
* be called from atomic context if sleep_us or timeout_us are used.
*/
#define readx_poll_timeout(op, addr, val, cond, sleep_us, timeout_us) \
read_poll_timeout(op, val, cond, sleep_us, timeout_us, false, addr)
/**
* readx_poll_timeout_atomic - Periodically poll an address until a condition is met or a timeout occurs
* @op: accessor function (takes @addr as its only argument)
* @addr: Address to poll
* @val: Variable to read the value into
* @cond: Break condition (usually involving @val)
* @delay_us: Time to udelay between reads in us (0 tight-loops). Please
* read udelay() function description for details and
* limitations.
* @timeout_us: Timeout in us, 0 means never timeout
*
* When available, you'll probably want to use one of the specialized
* macros defined below rather than this macro directly.
*
* Returns: 0 on success and -ETIMEDOUT upon a timeout. In either
* case, the last read value at @addr is stored in @val.
*/
#define readx_poll_timeout_atomic(op, addr, val, cond, delay_us, timeout_us) \
read_poll_timeout_atomic(op, val, cond, delay_us, timeout_us, false, addr)
#define readb_poll_timeout(addr, val, cond, delay_us, timeout_us) \
readx_poll_timeout(readb, addr, val, cond, delay_us, timeout_us)
#define readb_poll_timeout_atomic(addr, val, cond, delay_us, timeout_us) \
readx_poll_timeout_atomic(readb, addr, val, cond, delay_us, timeout_us)
#define readw_poll_timeout(addr, val, cond, delay_us, timeout_us) \
readx_poll_timeout(readw, addr, val, cond, delay_us, timeout_us)
#define readw_poll_timeout_atomic(addr, val, cond, delay_us, timeout_us) \
readx_poll_timeout_atomic(readw, addr, val, cond, delay_us, timeout_us)
#define readl_poll_timeout(addr, val, cond, delay_us, timeout_us) \
readx_poll_timeout(readl, addr, val, cond, delay_us, timeout_us)
#define readl_poll_timeout_atomic(addr, val, cond, delay_us, timeout_us) \
readx_poll_timeout_atomic(readl, addr, val, cond, delay_us, timeout_us)
#define readq_poll_timeout(addr, val, cond, delay_us, timeout_us) \
readx_poll_timeout(readq, addr, val, cond, delay_us, timeout_us)
#define readq_poll_timeout_atomic(addr, val, cond, delay_us, timeout_us) \
readx_poll_timeout_atomic(readq, addr, val, cond, delay_us, timeout_us)
#define readb_relaxed_poll_timeout(addr, val, cond, delay_us, timeout_us) \
readx_poll_timeout(readb_relaxed, addr, val, cond, delay_us, timeout_us)
#define readb_relaxed_poll_timeout_atomic(addr, val, cond, delay_us, timeout_us) \
readx_poll_timeout_atomic(readb_relaxed, addr, val, cond, delay_us, timeout_us)
#define readw_relaxed_poll_timeout(addr, val, cond, delay_us, timeout_us) \
readx_poll_timeout(readw_relaxed, addr, val, cond, delay_us, timeout_us)
#define readw_relaxed_poll_timeout_atomic(addr, val, cond, delay_us, timeout_us) \
readx_poll_timeout_atomic(readw_relaxed, addr, val, cond, delay_us, timeout_us)
#define readl_relaxed_poll_timeout(addr, val, cond, delay_us, timeout_us) \
readx_poll_timeout(readl_relaxed, addr, val, cond, delay_us, timeout_us)
#define readl_relaxed_poll_timeout_atomic(addr, val, cond, delay_us, timeout_us) \
readx_poll_timeout_atomic(readl_relaxed, addr, val, cond, delay_us, timeout_us)
#define readq_relaxed_poll_timeout(addr, val, cond, delay_us, timeout_us) \
readx_poll_timeout(readq_relaxed, addr, val, cond, delay_us, timeout_us)
#define readq_relaxed_poll_timeout_atomic(addr, val, cond, delay_us, timeout_us) \
readx_poll_timeout_atomic(readq_relaxed, addr, val, cond, delay_us, timeout_us)
#endif /* _LINUX_IOPOLL_H */