Regmap IRQ fix and related changes CPCAP
Subsystems (3)
v2 cover Regmap IRQ fix and related changes CPCAP
Hi all,
Here's v2 set of fixes to make CPCAP PMIC interrupts work reliably when used with multiple drivers. While working on the ADC, charger and USB PHY drivers I noticed that the PMIC interrupt to the SoC would eventually stop working.
All these can wait for v4.12 merge window as these issues don't show up currently. Two of the patches are also harmless currently, so I've kept them in this series to avoid pointless merge conflicts.
Regards,
Tony
Changes since v1:
- Updated regap-irq patch to use out_runtime_put in regmap_irq_thread also if pm_runtime_get() fails
- Clarify patch description for regmap-irq changes to make clear this is an issue with the CPCAP PMIC and not SoC GPIO edge/level handling based on comments from Charles Keepax <[email protected]>
- Collected acks
Tony Lindgren (4):
regmap: irq: Fix lost interrupts by introducing handle_reread
mfd: cpcap: Use handle_reread flag for interrupts
mfd: cpcap: Use ack_invert interrupts
mfd: cpcap: Fix bad use of IRQ sense register
drivers/base/regmap/regmap-irq.c | 77 +++++++++++++++++++++++++++------------- drivers/mfd/motorola-cpcap.c | 7 +++- include/linux/regmap.h | 2 ++ 3 files changed, 61 insertions(+), 25 deletions(-)
Signature · 1 line
2.11.1
PATCH v2 1/4 regmap: irq: Fix lost interrupts by introducing handle_reread
At least Motorola CPCAP PMIC needs it's device interrupts re-read until there are no more interrupts. Otherwise the PMIC interrupt to the SoC will eventually stop toggling. This seems to be a bug in the CPCAP PMIC where it can stop driving the GPIO interrupt to the SoC with pending CPCAP interrupts.
Let's allow handling issues like this by introducing a flag for handle_reread and by splitting regmap_irq_thread() into two separate functions for regmap_read_irq_status() and regmap_irq_handle_pending().
--- drivers/base/regmap/regmap-irq.c | 77 +++++++++++++++++++++++++++------------- include/linux/regmap.h | 2 ++ 2 files changed, 55 insertions(+), 24 deletions(-)
drivers/base/regmap/regmap-irq.c +53 -24
@@ -259,27 +259,11 @@ static const struct irq_chip regmap_irq_chip = { .irq_set_wake = regmap_irq_set_wake, }; -static irqreturn_t regmap_irq_thread (int irq,void *d) +static int regmap_read_irq_status (struct regmap_irq_chip_data *data) { -struct regmap_irq_chip_data *data = d;const struct regmap_irq_chip *chip = data->chip;struct regmap *map = data->map;int ret, i; -bool handled = false; -u32 reg; - -if (chip->handle_pre_irq) - chip->handle_pre_irq (chip->irq_drv_data); - -if (chip->runtime_pm) { - ret =pm_runtime_get_sync (map->dev); -if (ret <0 ) { -dev_err (map->dev,"IRQ thread failed to resume: %d\n" , - ret); -pm_runtime_put (map->dev); - goto exit; - } - }/* * Read in the statuses, using a single bulk read if possible
@@ -299,7 +283,7 @@ static irqreturn_t regmap_irq_thread(int irq, void *d)if (ret !=0 ) {dev_err (map->dev,"Failed to read IRQ status: %d\n" , ret); - goto exit; +return ret; }for (i =0 ; i < data->chip->num_regs; i++) {
@@ -315,7 +299,7 @@ static irqreturn_t regmap_irq_thread(int irq, void *d)break ;default :BUG (); - goto exit; +return ret; } }
@@ -330,13 +314,21 @@ static irqreturn_t regmap_irq_thread(int irq, void *d)dev_err (map->dev,"Failed to read IRQ status: %d\n" , ret); -if (chip->runtime_pm) -pm_runtime_put (map->dev); - goto exit; +return ret; } } } +return 0 ; +} + +static int regmap_irq_handle_pending (struct regmap_irq_chip_data *data) +{ +const struct regmap_irq_chip *chip = data->chip; +struct regmap *map = data->map; +int ret, i, handled =0 ; +u32 reg; +/* * Ignore masked IRQs and ack if we need to; we ack early so * there is no race between handling and acknowleding the
@@ -361,14 +353,51 @@ static irqreturn_t regmap_irq_thread(int irq, void *d)if (data->status_buf[chip->irqs[i].reg_offset / map->reg_stride] & chip->irqs[i].mask) {handle_nested_irq (irq_find_mapping (data->domain, i)); - handled = true; + handled++; + } + } + +return handled; +} + +static irqreturn_t regmap_irq_thread (int irq,void *d) +{ +struct regmap_irq_chip_data *data = d; +const struct regmap_irq_chip *chip = data->chip; +struct regmap *map = data->map; +int ret, handled =0 ; + +if (chip->handle_pre_irq) + chip->handle_pre_irq (chip->irq_drv_data); + +if (chip->runtime_pm) { + ret =pm_runtime_get_sync (map->dev); +if (ret <0 ) { +dev_err (map->dev,"IRQ thread failed to resume: %d\n" , + ret); + goto out_runtime_put; } } +do { + ret =regmap_read_irq_status (data); +if (ret) + goto out_runtime_put; + + ret =regmap_irq_handle_pending (data); +if (ret <0 ) + goto out_runtime_put; + +if (!ret) +break ; + + handled += ret; + }while (chip->handle_reread); + +out_runtime_put:if (chip->runtime_pm)pm_runtime_put (map->dev); -exit:if (chip->handle_post_irq) chip->handle_post_irq (chip->irq_drv_data);
include/linux/regmap.h +2 -0
@@ -897,6 +897,7 @@ struct regmap_irq {* @ack_invert: Inverted ack register: cleared bits for ack. * @wake_invert: Inverted wake register: cleared bits are wake enabled. * @type_invert: Invert the type flags. +* @handle_reread: Read interrupt status until no more interrupts are seen. * @runtime_pm: Hold a runtime PM lock on the device when accessing it. * * @num_regs: Number of registers in each control bank.
@@ -934,6 +935,7 @@ struct regmap_irq_chip { bool wake_invert:1; bool runtime_pm:1; bool type_invert:1; + bool handle_reread:1;int num_regs;
Signature · 1 line
2.11.1
PATCH v2 2/4 mfd: cpcap: Use handle_reread flag for interrupts
In reply to Tony LindgrenOn CPCAP we need to keep reading interrupts until there are no more interrupts. Otherwise the PMIC interrupt to the SoC will at some point stop toggling. This seems to happen because new CPCAP device interrupts show up while we're handling.
--- drivers/mfd/motorola-cpcap.c | 3 +++ 1 file changed, 3 insertions(+)
drivers/mfd/motorola-cpcap.c +3 -0
@@ -71,6 +71,7 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .ack_base = CPCAP_REG_MI1, .mask_base = CPCAP_REG_MIM1, .use_ack = true, + .handle_reread = true, }, { .name = "cpcap-m2",
@@ -79,6 +80,7 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .ack_base = CPCAP_REG_MI2, .mask_base = CPCAP_REG_MIM2, .use_ack = true, + .handle_reread = true, }, { .name = "cpcap1-4",
@@ -88,6 +90,7 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .mask_base = CPCAP_REG_INTM1, .type_base = CPCAP_REG_INTS1, .use_ack = true, + .handle_reread = true, }, };
Signature · 1 line
2.11.1
PATCH v2 3/4 mfd: cpcap: Use ack_invert interrupts
In reply to Tony LindgrenWe should use ack_invert as the int_read_and_clear() in the Motorola kernel tree does "ireg_val & ~mreg_val" before writing to the mask register.
--- drivers/mfd/motorola-cpcap.c | 3 +++ 1 file changed, 3 insertions(+)
drivers/mfd/motorola-cpcap.c +3 -0
@@ -72,6 +72,7 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .mask_base = CPCAP_REG_MIM1, .use_ack = true, .handle_reread = true, + .ack_invert = true, }, { .name = "cpcap-m2",
@@ -81,6 +82,7 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .mask_base = CPCAP_REG_MIM2, .use_ack = true, .handle_reread = true, + .ack_invert = true, }, { .name = "cpcap1-4",
@@ -91,6 +93,7 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .type_base = CPCAP_REG_INTS1, .use_ack = true, .handle_reread = true, + .ack_invert = true, }, };
Signature · 1 line
2.11.1
PATCH v2 4/4 mfd: cpcap: Fix bad use of IRQ sense register
In reply to Tony LindgrenThe cpcap INTS registers are for getting the value of the line, not for configuring the type.
--- drivers/mfd/motorola-cpcap.c | 1 - 1 file changed, 1 deletion(-)
drivers/mfd/motorola-cpcap.c +0 -1
@@ -90,7 +90,6 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .status_base = CPCAP_REG_INT1, .ack_base = CPCAP_REG_INT1, .mask_base = CPCAP_REG_INTM1, - .type_base = CPCAP_REG_INTS1, .use_ack = true, .handle_reread = true, .ack_invert = true,
Signature · 1 line
2.11.1
Mark Brown
maintainerRe: [PATCH 1/4] regmap: irq: Fix lost interrupts by introducing handle_reread
In reply to Tony LindgrenQuoted 9 lines · Tony Lindgren
At least Motorola CPCAP PMIC needs it's device interrupts re-read until there are no more interrupts. Otherwise the PMIC interrupt to the SoC will eventually stop toggling. This seems to be a bug in the CPCAP PMIC where it can stop driving the GPIO interrupt to the SoC with pending CPCAP interrupts.
Let's allow handling issues like this by introducing a flag for
handle_rereadand by splittingregmap_irq_thread()into two separate functions forregmap_read_irq_status()andregmap_irq_handle_pending().
So, I see your use case but the fact is that as Charles observed this is exactly the code used for emulating level triggered IRQs with edge triggered interrupt controllers. This means someone is doubtless going to end up using it for precisely that. This makes me uncomfortable, we do have this open coded into various drivers already but this is more of a core thing and it feels like this should be in genirq rather than here... that said, looking at the code:
Quoted 4 lines
drivers/base/regmap/regmap-irq.c +4
-0+ do { + ret = regmap_read_irq_status(data); + if (ret) + goto out_runtime_put;
drivers/base/regmap/regmap-irq.c static irqreturn_t regmap_irq_thread(int irq, void *d) · line 395
Quoted 10 lines
drivers/base/regmap/regmap-irq.c +10
-0+ + ret =regmap_irq_handle_pending (data); +if (ret <0 ) + goto out_runtime_put; + +if (!ret) +break ; + + handled += ret; + }while (chip->handle_reread);
There's no protection against screaming interrupts here, I'd really like to see that. Also some tracing of the number of times we spin.
Re: [PATCH 1/4] regmap: irq: Fix lost interrupts by introducing handle_reread
Quoted 11 lines · Mark Brown
Quoted 9 lines · Tony Lindgren
At least Motorola CPCAP PMIC needs it's device interrupts re-read until there are no more interrupts. Otherwise the PMIC interrupt to the SoC will eventually stop toggling. This seems to be a bug in the CPCAP PMIC where it can stop driving the GPIO interrupt to the SoC with pending CPCAP interrupts.
Let's allow handling issues like this by introducing a flag for
handle_rereadand by splittingregmap_irq_thread()into two separate functions forregmap_read_irq_status()andregmap_irq_handle_pending().
Quoted 7 lines
So, I see your use case but the fact is that as Charles observed this is exactly the code used for emulating level triggered IRQs with edge triggered interrupt controllers. This means someone is doubtless going to end up using it for precisely that. This makes me uncomfortable, we do have this open coded into various drivers already but this is more of a core thing and it feels like this should be in genirq rather than here... that said, looking at the code:
Yes.. But then again we might avoid piling up yet more driver specific hacks. I don't know what the genirq solution would look like, handle until we get IRQ_NONE? :)
Quoted 9 lines
Quoted 9 lines
(quoted diff) +9
-0+ do { + ret = regmap_read_irq_status(data); + if (ret) + goto out_runtime_put; + + ret = regmap_irq_handle_pending(data); + if (ret < 0) + goto out_runtime_put; +
Quoted 8 lines
Review
Quoted 5 lines
(quoted diff) +5
-0+ if (!ret) + break; + + handled += ret; + } while (chip->handle_reread);There's no protection against screaming interrupts here, I'd really like to see that. Also some tracing of the number of times we spin.
Good idea. How about something like below where handle_reread checks for the total time spent in the thread loop with a large enough timeout? Or do you have some better ideas in mind?
I tested I can hit that warning with timeout set to much lower 10 ms with retries being 1 or 2 at that point.
Regards,
Tony
8< -------------------------------
Quoted 1 line
From tony Mon Sep 17 00:00:00 2001
From: Tony Lindgren <[email protected]>
Date: Mon, 27 Mar 2017 08:09:36 -0700
Subject: [PATCH] regmap: irq: Fix lost interrupts by introducing
handle_reread
At least Motorola CPCAP PMIC needs it's device interrupts re-read until there are no more interrupts. Otherwise the PMIC interrupt to the SoC will eventually stop toggling. This seems to be a bug in the CPCAP PMIC where it can stop driving the GPIO interrupt to the SoC with pending CPCAP interrupts.
Let's allow handling issues like this by introducing a flag for handle_reread and by splitting regmap_irq_thread() into two separate functions for regmap_read_irq_status() and regmap_irq_handle_pending().
--- drivers/base/regmap/regmap-irq.c | 97 ++++++++++++++++++++++++++++++---------- include/linux/regmap.h | 2 + 2 files changed, 75 insertions(+), 24 deletions(-)
drivers/base/regmap/regmap-irq.c +73 -24
@@ -21,6 +21,8 @@#include "internal.h" +#define REGMAP_IRQ_THREAD_MAX_MSECS 3000 +struct regmap_irq_chip_data {struct mutex lock;struct irq_chip irq_chip;
@@ -259,27 +261,11 @@ static const struct irq_chip regmap_irq_chip = { .irq_set_wake = regmap_irq_set_wake, }; -static irqreturn_t regmap_irq_thread (int irq,void *d) +static int regmap_read_irq_status (struct regmap_irq_chip_data *data) { -struct regmap_irq_chip_data *data = d;const struct regmap_irq_chip *chip = data->chip;struct regmap *map = data->map;int ret, i; -bool handled = false; -u32 reg; - -if (chip->handle_pre_irq) - chip->handle_pre_irq (chip->irq_drv_data); - -if (chip->runtime_pm) { - ret =pm_runtime_get_sync (map->dev); -if (ret <0 ) { -dev_err (map->dev,"IRQ thread failed to resume: %d\n" , - ret); -pm_runtime_put (map->dev); - goto exit; - } - }/* * Read in the statuses, using a single bulk read if possible
@@ -299,7 +285,7 @@ static irqreturn_t regmap_irq_thread(int irq, void *d)if (ret !=0 ) {dev_err (map->dev,"Failed to read IRQ status: %d\n" , ret); - goto exit; +return ret; }for (i =0 ; i < data->chip->num_regs; i++) {
@@ -315,7 +301,7 @@ static irqreturn_t regmap_irq_thread(int irq, void *d)break ;default :BUG (); - goto exit; +return ret; } }
@@ -330,13 +316,21 @@ static irqreturn_t regmap_irq_thread(int irq, void *d)dev_err (map->dev,"Failed to read IRQ status: %d\n" , ret); -if (chip->runtime_pm) -pm_runtime_put (map->dev); - goto exit; +return ret; } } } +return 0 ; +} + +static int regmap_irq_handle_pending (struct regmap_irq_chip_data *data) +{ +const struct regmap_irq_chip *chip = data->chip; +struct regmap *map = data->map; +int ret, i, handled =0 ; +u32 reg; +/* * Ignore masked IRQs and ack if we need to; we ack early so * there is no race between handling and acknowleding the
@@ -361,14 +355,69 @@ static irqreturn_t regmap_irq_thread(int irq, void *d)if (data->status_buf[chip->irqs[i].reg_offset / map->reg_stride] & chip->irqs[i].mask) {handle_nested_irq (irq_find_mapping (data->domain, i)); - handled = true; + handled++; } } +return handled; +} + +static irqreturn_t regmap_irq_thread (int irq,void *d) +{ +struct regmap_irq_chip_data *data = d; +const struct regmap_irq_chip *chip = data->chip; +struct regmap *map = data->map; +ktime_t calltime =0 , retrytime; +int ret, handled =0 , retries =0 ; + +if (chip->handle_reread) + calltime =ktime_get (); + +if (chip->handle_pre_irq) + chip->handle_pre_irq (chip->irq_drv_data); + +if (chip->runtime_pm) { + ret =pm_runtime_get_sync (map->dev); +if (ret <0 ) { +dev_err (map->dev,"IRQ thread failed to resume: %d\n" , + ret); + goto out_runtime_put; + } + } + +do { + ret =regmap_read_irq_status (data); +if (ret) + goto out_runtime_put; + + ret =regmap_irq_handle_pending (data); +if (ret <0 ) + goto out_runtime_put; + +if (!ret) +break ; + + handled += ret; + retries++; + +if (chip->handle_reread) { +s64 msecs; + + retrytime =ktime_get (); + msecs =ktime_to_ms (ktime_sub (retrytime, calltime)); +if (msecs >REGMAP_IRQ_THREAD_MAX_MSECS ) { +dev_err (map->dev, +"IRQ thread timeout after %i retries\n" , + retries); + goto out_runtime_put; + } + } + }while (chip->handle_reread); + +out_runtime_put:if (chip->runtime_pm)pm_runtime_put (map->dev); -exit:if (chip->handle_post_irq) chip->handle_post_irq (chip->irq_drv_data);
include/linux/regmap.h +2 -0
@@ -897,6 +897,7 @@ struct regmap_irq {* @ack_invert: Inverted ack register: cleared bits for ack. * @wake_invert: Inverted wake register: cleared bits are wake enabled. * @type_invert: Invert the type flags. +* @handle_reread: Read interrupt status until no more interrupts are seen. * @runtime_pm: Hold a runtime PM lock on the device when accessing it. * * @num_regs: Number of registers in each control bank.
@@ -934,6 +935,7 @@ struct regmap_irq_chip { bool wake_invert:1; bool runtime_pm:1; bool type_invert:1; + bool handle_reread:1;int num_regs;
Signature · 1 line
2.12.1
Mark Brown
maintainerRe: [PATCH 1/4] regmap: irq: Fix lost interrupts by introducing handle_reread
Quoted 10 lines · Tony Lindgren
Quoted 7 lines · Mark Brown
So, I see your use case but the fact is that as Charles observed this is exactly the code used for emulating level triggered IRQs with edge triggered interrupt controllers. This means someone is doubtless going to end up using it for precisely that. This makes me uncomfortable, we do have this open coded into various drivers already but this is more of a core thing and it feels like this should be in genirq rather than here... that said, looking at the code:
Quoted 2 lines
Yes.. But then again we might avoid piling up yet more driver specific hacks. I don't know what the genirq solution would look
Right, my thinking here is that by pushing into genirq we minimise the need even further since it'll also be available to drivers not using regmap-irq.
Quoted 1 line
like, handle until we get
IRQ_NONE? :)
Well, that's what the per driver emulation does so... yeah. Probably with an upper limit on the number of times we do that.
Quoted 9 lines
Quoted 2 lines
There's no protection against screaming interrupts here, I'd really like to see that. Also some tracing of the number of times we spin.
Good idea. How about something like below where
handle_rereadchecks for the total time spent in the thread loop with a large enough timeout? Or do you have some better ideas in mind?I tested I can hit that warning with timeout set to much lower 10 ms with retries being 1 or 2 at that point.
That looks sensible, yes.
Re: [PATCH 1/4] regmap: irq: Fix lost interrupts by introducing handle_reread
Quoted 14 lines · Mark Brown
Quoted 10 lines · Tony Lindgren
Quoted 7 lines · Mark Brown
So, I see your use case but the fact is that as Charles observed this is exactly the code used for emulating level triggered IRQs with edge triggered interrupt controllers. This means someone is doubtless going to end up using it for precisely that. This makes me uncomfortable, we do have this open coded into various drivers already but this is more of a core thing and it feels like this should be in genirq rather than here... that said, looking at the code:
Quoted 2 lines
Yes.. But then again we might avoid piling up yet more driver specific hacks. I don't know what the genirq solution would look
Quoted 8 lines
Right, my thinking here is that by pushing into genirq we minimise the need even further since it'll also be available to drivers not using regmap-irq.
Quoted 1 line
like, handle until we get
IRQ_NONE? :)Well, that's what the per driver emulation does so... yeah. Probably with an upper limit on the number of times we do that.
OK let's first see how that would work. I'll send a patch for that.
Quoted 11 lines
Quoted 9 lines
Quoted 2 lines
There's no protection against screaming interrupts here, I'd really like to see that. Also some tracing of the number of times we spin.
Good idea. How about something like below where
handle_rereadchecks for the total time spent in the thread loop with a large enough timeout? Or do you have some better ideas in mind?I tested I can hit that warning with timeout set to much lower 10 ms with retries being 1 or 2 at that point.
That looks sensible, yes.
OK
Regards,
Tony
Mark Brown
maintainerRe: [PATCH 1/4] regmap: irq: Fix lost interrupts by introducing handle_reread
Quoted 6 lines · Tony Lindgren
Quoted 3 lines · Mark Brown
Right, my thinking here is that by pushing into genirq we minimise the need even further since it'll also be available to drivers not using regmap-irq.
Quoted 7 lines
Quoted 4 lines
Quoted 1 line
like, handle until we get
IRQ_NONE? :)Well, that's what the per driver emulation does so... yeah. Probably with an upper limit on the number of times we do that.
OK let's first see how that would work. I'll send a patch for that.
Thanks. Can you keep me on the CC please? It's something I keep thinking about looking at myself.
Re: [PATCH 1/4] regmap: irq: Fix lost interrupts by introducing handle_reread
Quoted 9 lines · Mark Brown
Quoted 7 lines · Tony Lindgren
Quoted 5 lines · Mark Brown
Right, my thinking here is that by pushing into genirq we minimise the need even further since it'll also be available to drivers not using regmap-irq.
Quoted 1 line
like, handle until we get
IRQ_NONE? :)
Quoted 8 lines
Quoted 5 lines
Quoted 2 lines
Well, that's what the per driver emulation does so... yeah. Probably with an upper limit on the number of times we do that.
OK let's first see how that would work. I'll send a patch for that.
Thanks. Can you keep me on the CC please? It's something I keep thinking about looking at myself.
Sure will do, you'll get some shared flames on it :)
Regards,
Tony
Lee Jones
AppliedRe: [PATCH 2/4] mfd: cpcap: Use handle_reread flag for interrupts
In reply to Tony LindgrenQuoted 11 lines · Tony Lindgren
On CPCAP we need to keep reading interrupts until there are no more interrupts. Otherwise the PMIC interrupt to the SoC will at some point stop toggling. This seems to happen because new CPCAP device interrupts show up while we're handling.
Cc Charles Keepax <[email protected]>Cc Lee Jones <[email protected]>Cc Marcel Partap <[email protected]>Cc Michael Scott <[email protected]>Tested-by Sebastian Reichel <[email protected]>Signed-off-by Tony Lindgren <[email protected]>
Quoted 3 lines
--- drivers/mfd/motorola-cpcap.c | 3 +++ 1 file changed, 3 insertions(+)
Applied, thanks.
Quoted 27 lines
drivers/mfd/motorola-cpcap.c +3
-0@@ -71,6 +71,7 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .ack_base = CPCAP_REG_MI1, .mask_base = CPCAP_REG_MIM1, .use_ack = true, + .handle_reread = true, }, { .name = "cpcap-m2",@@ -79,6 +80,7 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .ack_base = CPCAP_REG_MI2, .mask_base = CPCAP_REG_MIM2, .use_ack = true, + .handle_reread = true, }, { .name = "cpcap1-4",@@ -88,6 +90,7 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .mask_base = CPCAP_REG_INTM1, .type_base = CPCAP_REG_INTS1, .use_ack = true, + .handle_reread = true, }, };
Signature · 4 lines
Lee Jones Linaro STMicroelectronics Landing Team Lead Linaro.org │ Open source software for ARM SoCs Follow Linaro: Facebook | Twitter | Blog
Lee Jones
AppliedRe: [PATCH 3/4] mfd: cpcap: Use ack_invert interrupts
In reply to Tony LindgrenQuoted 10 lines · Tony Lindgren
We should use
ack_invertas theint_read_and_clear()in the Motorola kernel tree does "ireg_val& ~mreg_val" before writing to the mask register.Cc Charles Keepax <[email protected]>Cc Lee Jones <[email protected]>Cc Marcel Partap <[email protected]>Cc Michael Scott <[email protected]>Tested-by Sebastian Reichel <[email protected]>Signed-off-by Tony Lindgren <[email protected]>
Quoted 3 lines
--- drivers/mfd/motorola-cpcap.c | 3 +++ 1 file changed, 3 insertions(+)
Applied, thanks.
Quoted 27 lines
drivers/mfd/motorola-cpcap.c +3
-0@@ -72,6 +72,7 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .mask_base = CPCAP_REG_MIM1, .use_ack = true, .handle_reread = true, + .ack_invert = true, }, { .name = "cpcap-m2",@@ -81,6 +82,7 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .mask_base = CPCAP_REG_MIM2, .use_ack = true, .handle_reread = true, + .ack_invert = true, }, { .name = "cpcap1-4",@@ -91,6 +93,7 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .type_base = CPCAP_REG_INTS1, .use_ack = true, .handle_reread = true, + .ack_invert = true, }, };
Signature · 4 lines
Lee Jones Linaro STMicroelectronics Landing Team Lead Linaro.org │ Open source software for ARM SoCs Follow Linaro: Facebook | Twitter | Blog
Lee Jones
AppliedRe: [PATCH 4/4] mfd: cpcap: Fix bad use of IRQ sense register
In reply to Tony LindgrenQuoted 12 lines · Tony Lindgren
The cpcap INTS registers are for getting the value of the line, not for configuring the type.
Cc Charles Keepax <[email protected]>Cc Lee Jones <[email protected]>Cc Marcel Partap <[email protected]>Cc Michael Scott <[email protected]>Cc Sebastian Reichel <[email protected]>Reviewed-by Sebastian Reichel <[email protected]>Tested-by Sebastian Reichel <[email protected]>Acked-by Lee Jones <[email protected]>Signed-off-by Tony Lindgren <[email protected]>
Quoted 3 lines
--- drivers/mfd/motorola-cpcap.c | 1 - 1 file changed, 1 deletion(-)
Applied, thanks.
Quoted 11 lines
drivers/mfd/motorola-cpcap.c +0
-1@@ -90,7 +90,6 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .status_base = CPCAP_REG_INT1, .ack_base = CPCAP_REG_INT1, .mask_base = CPCAP_REG_INTM1, - .type_base = CPCAP_REG_INTS1, .use_ack = true, .handle_reread = true, .ack_invert = true,
Signature · 4 lines
Lee Jones Linaro STMicroelectronics Landing Team Lead Linaro.org │ Open source software for ARM SoCs Follow Linaro: Facebook | Twitter | Blog
Re: [PATCH 1/4] regmap: irq: Fix lost interrupts by introducing handle_reread
In reply to Tony LindgrenQuoted 16 lines · Tony Lindgren
At least Motorola CPCAP PMIC needs it's device interrupts re-read until there are no more interrupts. Otherwise the PMIC interrupt to the SoC will eventually stop toggling. This seems to be a bug in the CPCAP PMIC where it can stop driving the GPIO interrupt to the SoC with pending CPCAP interrupts.
Let's allow handling issues like this by introducing a flag for
handle_rereadand by splittingregmap_irq_thread()into two separate functions forregmap_read_irq_status()andregmap_irq_handle_pending().Cc Charles Keepax <[email protected]>Cc Lee Jones <[email protected]>Cc Marcel Partap <[email protected]>Cc Michael Scott <[email protected]>Tested-by Sebastian Reichel <[email protected]>Signed-off-by Tony Lindgren <[email protected]>
Quoted 4 lines
--- drivers/base/regmap/regmap-irq.c | 77 +++++++++++++++++++++++++++------------- include/linux/regmap.h | 2 ++ 2 files changed, 55 insertions(+), 24 deletions(-)
Looks like this is required for the MFD patches.
I'm removing them for now.
Quoted 148 lines
drivers/base/regmap/regmap-irq.c +53
-24@@ -259,27 +259,11 @@ static const struct irq_chip regmap_irq_chip = { .irq_set_wake = regmap_irq_set_wake, }; -static irqreturn_t regmap_irq_thread (int irq,void *d) +static int regmap_read_irq_status (struct regmap_irq_chip_data *data) { -struct regmap_irq_chip_data *data = d;const struct regmap_irq_chip *chip = data->chip;struct regmap *map = data->map;int ret, i; -bool handled = false; -u32 reg; - -if (chip->handle_pre_irq) - chip->handle_pre_irq (chip->irq_drv_data); - -if (chip->runtime_pm) { - ret =pm_runtime_get_sync (map->dev); -if (ret <0 ) { -dev_err (map->dev,"IRQ thread failed to resume: %d\n" , - ret); -pm_runtime_put (map->dev); - goto exit; - } - }/* * Read in the statuses, using a single bulk read if possible @@ -299,7 +283,7 @@ static irqreturn_t regmap_irq_thread(int irq, void *d)if (ret !=0 ) {dev_err (map->dev,"Failed to read IRQ status: %d\n" , ret); - goto exit; +return ret; }for (i =0 ; i < data->chip->num_regs; i++) {@@ -315,7 +299,7 @@ static irqreturn_t regmap_irq_thread(int irq, void *d)break ;default :BUG (); - goto exit; +return ret; } }@@ -330,13 +314,21 @@ static irqreturn_t regmap_irq_thread(int irq, void *d)dev_err (map->dev,"Failed to read IRQ status: %d\n" , ret); -if (chip->runtime_pm) -pm_runtime_put (map->dev); - goto exit; +return ret; } } } +return 0 ; +} + +static int regmap_irq_handle_pending (struct regmap_irq_chip_data *data) +{ +const struct regmap_irq_chip *chip = data->chip; +struct regmap *map = data->map; +int ret, i, handled =0 ; +u32 reg; +/* * Ignore masked IRQs and ack if we need to; we ack early so * there is no race between handling and acknowleding the @@ -361,14 +353,51 @@ static irqreturn_t regmap_irq_thread(int irq, void *d)if (data->status_buf[chip->irqs[i].reg_offset / map->reg_stride] & chip->irqs[i].mask) {handle_nested_irq (irq_find_mapping (data->domain, i)); - handled = true; + handled++; + } + } + +return handled; +} + +static irqreturn_t regmap_irq_thread (int irq,void *d) +{ +struct regmap_irq_chip_data *data = d; +const struct regmap_irq_chip *chip = data->chip; +struct regmap *map = data->map; +int ret, handled =0 ; + +if (chip->handle_pre_irq) + chip->handle_pre_irq (chip->irq_drv_data); + +if (chip->runtime_pm) { + ret =pm_runtime_get_sync (map->dev); +if (ret <0 ) { +dev_err (map->dev,"IRQ thread failed to resume: %d\n" , + ret); + goto out_runtime_put; } } +do { + ret =regmap_read_irq_status (data); +if (ret) + goto out_runtime_put; + + ret =regmap_irq_handle_pending (data); +if (ret <0 ) + goto out_runtime_put; + +if (!ret) +break ; + + handled += ret; + }while (chip->handle_reread); + +out_runtime_put:if (chip->runtime_pm)pm_runtime_put (map->dev); -exit:if (chip->handle_post_irq) chip->handle_post_irq (chip->irq_drv_data);include/linux/regmap.h +2
-0@@ -897,6 +897,7 @@ struct regmap_irq {* @ack_invert: Inverted ack register: cleared bits for ack. * @wake_invert: Inverted wake register: cleared bits are wake enabled. * @type_invert: Invert the type flags. +* @handle_reread: Read interrupt status until no more interrupts are seen. * @runtime_pm: Hold a runtime PM lock on the device when accessing it. * * @num_regs: Number of registers in each control bank. @@ -934,6 +935,7 @@ struct regmap_irq_chip { bool wake_invert:1; bool runtime_pm:1; bool type_invert:1; + bool handle_reread:1;int num_regs;
Signature · 4 lines
Lee Jones Linaro STMicroelectronics Landing Team Lead Linaro.org │ Open source software for ARM SoCs Follow Linaro: Facebook | Twitter | Blog
Re: [PATCH 2/4] mfd: cpcap: Use handle_reread flag for interrupts
In reply to Lee JonesHi,
Quoted 17 lines · Lee Jones
Quoted 11 lines · Tony Lindgren
On CPCAP we need to keep reading interrupts until there are no more interrupts. Otherwise the PMIC interrupt to the SoC will at some point stop toggling. This seems to happen because new CPCAP device interrupts show up while we're handling.
Cc Charles Keepax <[email protected]>Cc Lee Jones <[email protected]>Cc Marcel Partap <[email protected]>Cc Michael Scott <[email protected]>Tested-by Sebastian Reichel <[email protected]>Signed-off-by Tony Lindgren <[email protected]>Quoted 3 lines
--- drivers/mfd/motorola-cpcap.c | 3 +++ 1 file changed, 3 insertions(+)
Quoted 1 line
Applied, thanks.
Sorry this one depends on patch 1/4 for .handle_reread. So this one should be reverted or dropped. That will cause trivial merge conflicts with patches 2 and 4.
Please let me know if you want me to resend just patches 2 and 4 as patches 1 and 2 need more work.
Regards,
Tony
Quoted 33 lines
Quoted 27 lines
drivers/mfd/motorola-cpcap.c +3
-0@@ -71,6 +71,7 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .ack_base = CPCAP_REG_MI1, .mask_base = CPCAP_REG_MIM1, .use_ack = true, + .handle_reread = true, }, { .name = "cpcap-m2",@@ -79,6 +80,7 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .ack_base = CPCAP_REG_MI2, .mask_base = CPCAP_REG_MIM2, .use_ack = true, + .handle_reread = true, }, { .name = "cpcap1-4",@@ -88,6 +90,7 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .mask_base = CPCAP_REG_INTM1, .type_base = CPCAP_REG_INTS1, .use_ack = true, + .handle_reread = true, }, };Signature · 4 lines
Lee Jones Linaro STMicroelectronics Landing Team Lead Linaro.org │ Open source software for ARM SoCs Follow Linaro: Facebook | Twitter | Blog
Re: [PATCH 2/4] mfd: cpcap: Use handle_reread flag for interrupts
Quoted 20 lines · Tony Lindgren
Hi,
Quoted 17 lines · Lee Jones
Quoted 14 lines · Tony Lindgren
On CPCAP we need to keep reading interrupts until there are no more interrupts. Otherwise the PMIC interrupt to the SoC will at some point stop toggling. This seems to happen because new CPCAP device interrupts show up while we're handling.
Cc Charles Keepax <[email protected]>Cc Lee Jones <[email protected]>Cc Marcel Partap <[email protected]>Cc Michael Scott <[email protected]>Tested-by Sebastian Reichel <[email protected]>Signed-off-by Tony Lindgren <[email protected]>--- drivers/mfd/motorola-cpcap.c | 3 +++ 1 file changed, 3 insertions(+)
Quoted 8 lines
Quoted 1 line
Applied, thanks.
Sorry this one depends on patch 1/4 for .handle_reread. So this one should be reverted or dropped. That will cause trivial merge conflicts with patches 2 and 4.
Please let me know if you want me to resend just patches 2 and 4 as patches 1 and 2 need more work.
Yes, I found that out (and replied to the regmap patch).
I'll revert them all for now.
Awaiting a subsequent revision.
Quoted 28 lines
Quoted 28 lines
Quoted 27 lines
drivers/mfd/motorola-cpcap.c +3
-0@@ -71,6 +71,7 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .ack_base = CPCAP_REG_MI1, .mask_base = CPCAP_REG_MIM1, .use_ack = true, + .handle_reread = true, }, { .name = "cpcap-m2",@@ -79,6 +80,7 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .ack_base = CPCAP_REG_MI2, .mask_base = CPCAP_REG_MIM2, .use_ack = true, + .handle_reread = true, }, { .name = "cpcap1-4",@@ -88,6 +90,7 @@ static struct regmap_irq_chip cpcap_irq_chip[CPCAP_NR_IRQ_CHIPS] = { .mask_base = CPCAP_REG_INTM1, .type_base = CPCAP_REG_INTS1, .use_ack = true, + .handle_reread = true, }, };
Signature · 4 lines
Lee Jones Linaro STMicroelectronics Landing Team Lead Linaro.org │ Open source software for ARM SoCs Follow Linaro: Facebook | Twitter | Blog
Re: [PATCH 1/4] regmap: irq: Fix lost interrupts by introducing handle_reread
In reply to Tony LindgrenHi,
Quoted 13 lines · Tony Lindgren
Quoted 12 lines · Mark Brown
Quoted 10 lines · Tony Lindgren
Quoted 8 lines · Mark Brown
Right, my thinking here is that by pushing into genirq we minimise the need even further since it'll also be available to drivers not using regmap-irq.
Quoted 1 line
like, handle until we get
IRQ_NONE? :)Well, that's what the per driver emulation does so... yeah. Probably with an upper limit on the number of times we do that.
Quoted 7 lines
Quoted 5 lines
Quoted 2 lines
OK let's first see how that would work. I'll send a patch for that.
Thanks. Can you keep me on the CC please? It's something I keep thinking about looking at myself.
Sure will do, you'll get some shared flames on it :)
So I found the real problem after thinking what you guys commented on the "level device connected to an edge only GPIO". I added some printks to verify that's not the case just to find out that the dts GPIO interrupt edge configuration got only passed to the SPI driver :)
So the level IRQ changes I did earlier to the dts file did nothing.
The fix is simply to call devm_regmap_add_irq_chip() with irq_get_trigger_type(cpcap->spi->irq) to get the configured triggering passed properly from the dts.
So I'll drop the genirq/regmap_irq related hacks and resend just the minimal MFD fixes. Similar misconfiguration may be the root cause for other drivers too..
Regards,
Tony
Re: [PATCH 2/4] mfd: cpcap: Use handle_reread flag for interrupts
In reply to Lee JonesQuoted 28 lines · Lee Jones
Quoted 26 lines · Tony Lindgren
Hi,
Quoted 17 lines · Lee Jones
Quoted 11 lines · Tony Lindgren
On CPCAP we need to keep reading interrupts until there are no more interrupts. Otherwise the PMIC interrupt to the SoC will at some point stop toggling. This seems to happen because new CPCAP device interrupts show up while we're handling.
Cc Charles Keepax <[email protected]>Cc Lee Jones <[email protected]>Cc Marcel Partap <[email protected]>Cc Michael Scott <[email protected]>Tested-by Sebastian Reichel <[email protected]>Signed-off-by Tony Lindgren <[email protected]>Quoted 3 lines
--- drivers/mfd/motorola-cpcap.c | 3 +++ 1 file changed, 3 insertions(+)Quoted 1 line
Applied, thanks.
Sorry this one depends on patch 1/4 for .handle_reread. So this one should be reverted or dropped. That will cause trivial merge conflicts with patches 2 and 4.
Quoted 6 lines
Quoted 2 lines
Please let me know if you want me to resend just patches 2 and 4 as patches 1 and 2 need more work.
Yes, I found that out (and replied to the regmap patch).
I'll revert them all for now.
OK thanks.
Quoted 1 line
Awaiting a subsequent revision.
Sent as "[PATCHv3 0/3] CPCAP PMIC IRQ fix and related changes" with regmap changes dropped as I figured out what the real issue with lost interrupts was :)
Regards,
Tony
Mark Brown
maintainerRe: [PATCH 1/4] regmap: irq: Fix lost interrupts by introducing handle_reread
In reply to Tony LindgrenQuoted 3 lines · Tony Lindgren
So I'll drop the genirq/regmap_irq related hacks and resend just the minimal MFD fixes. Similar misconfiguration may be the root cause for other drivers too..
It is sadly far too common for people to implement interrupt controllers that only do edge triggers, I've no idea why even on what are supposed to be relatively high end SoCs. It seems to be a hardware designer thing, AFAICT they think for something to be useful it needs to be a bit more complicated.
Re: [PATCH 1/4] regmap: irq: Fix lost interrupts by introducing handle_reread
Quoted 11 lines · Mark Brown
Quoted 3 lines · Tony Lindgren
So I'll drop the genirq/regmap_irq related hacks and resend just the minimal MFD fixes. Similar misconfiguration may be the root cause for other drivers too..
It is sadly far too common for people to implement interrupt controllers that only do edge triggers, I've no idea why even on what are supposed to be relatively high end SoCs. It seems to be a hardware designer thing, AFAICT they think for something to be useful it needs to be a bit more complicated.
For edge only GPIO controllers handling level interrupts might be somewhat fixable in software. The GPIO controller could have a loop reading of the GPIO line status in the interrupt handler and comparing it to the configured triggering.
Regards,
Tony
Mark Brown
maintainerRe: [PATCH 1/4] regmap: irq: Fix lost interrupts by introducing handle_reread
Quoted 8 lines · Tony Lindgren
Quoted 5 lines · Mark Brown
It is sadly far too common for people to implement interrupt controllers that only do edge triggers, I've no idea why even on what are supposed to be relatively high end SoCs. It seems to be a hardware designer thing, AFAICT they think for something to be useful it needs to be a bit more complicated.
Quoted 4 lines
For edge only GPIO controllers handling level interrupts might be somewhat fixable in software. The GPIO controller could have a loop reading of the GPIO line status in the interrupt handler and comparing it to the configured triggering.
Yeah, it's doable if annoying.