i2c/busses: Avoid i2c interrupt status clear race condition.
ryan_chen
maintainerPATCH 1/1 i2c/busses: Avoid i2c interrupt status clear race condition.
In AST2600 there have a slow peripheral bus between CPU
and i2c controller.
Therefore GIC i2c interrupt status clear have delay timing, when CPU issue write clear i2c controller interrupt status. To avoid this issue, the driver need have read after write
clear at i2c ISR.
--- drivers/i2c/busses/i2c-aspeed.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/i2c/busses/i2c-aspeed.c +4 -1
@@ -603,6 +603,7 @@ static irqreturn_t aspeed_i2c_bus_irq(int irq, void *dev_id)/* Ack all interrupts except for Rx done */ writel (irq_received & ~ASPEED_I2CD_INTR_RX_DONE , bus->base +ASPEED_I2C_INTR_STS_REG ); +readl (bus->base +ASPEED_I2C_INTR_STS_REG ); irq_remaining = irq_received;#if IS_ENABLED (CONFIG_I2C_SLAVE )
@@ -645,9 +646,11 @@ static irqreturn_t aspeed_i2c_bus_irq(int irq, void *dev_id) irq_received, irq_handled);/* Ack Rx done */ -if (irq_received &ASPEED_I2CD_INTR_RX_DONE ) +if (irq_received &ASPEED_I2CD_INTR_RX_DONE ) {writel (ASPEED_I2CD_INTR_RX_DONE , bus->base +ASPEED_I2C_INTR_STS_REG ); +readl (bus->base +ASPEED_I2C_INTR_STS_REG ); + }spin_unlock (&bus->lock);return irq_remaining ?IRQ_NONE :IRQ_HANDLED ; }
Signature · 1 line
2.17.1
Quoted 8 lines · ryan_chen
In AST2600 there have a slow peripheral bus between CPU
and i2c controller.
Therefore GIC i2c interrupt status clear have delay timing, when CPU issue write clear i2c controller interrupt status. To avoid this issue, the driver need have read after write
clear at i2c ISR.Signed-off-by ryan_chen <[email protected]>
v0? is it a prototype?
And is there maybe a Fixes: tag for it?
Quoted 32 lines
--- drivers/i2c/busses/i2c-aspeed.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)drivers/i2c/busses/i2c-aspeed.c +4
-1@@ -603,6 +603,7 @@ static irqreturn_t aspeed_i2c_bus_irq(int irq, void *dev_id)/* Ack all interrupts except for Rx done */ writel (irq_received & ~ASPEED_I2CD_INTR_RX_DONE , bus->base +ASPEED_I2C_INTR_STS_REG ); +readl (bus->base +ASPEED_I2C_INTR_STS_REG ); irq_remaining = irq_received;#if IS_ENABLED (CONFIG_I2C_SLAVE )@@ -645,9 +646,11 @@ static irqreturn_t aspeed_i2c_bus_irq(int irq, void *dev_id) irq_received, irq_handled);/* Ack Rx done */ -if (irq_received &ASPEED_I2CD_INTR_RX_DONE ) +if (irq_received &ASPEED_I2CD_INTR_RX_DONE ) {writel (ASPEED_I2CD_INTR_RX_DONE , bus->base +ASPEED_I2C_INTR_STS_REG ); +readl (bus->base +ASPEED_I2C_INTR_STS_REG ); + }spin_unlock (&bus->lock);return irq_remaining ?IRQ_NONE :IRQ_HANDLED ; }Signature · 1 line
2.17.1
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://lists.ozlabs.org/pipermail/linux-aspeed/attachments/20200429/b98af618/attachment.sig>
Ryan Chen
maintainer-----Original Message-----
From: Wolfram Sang [mailto:wsa at the-dreams.de]
Sent: Wednesday, April 29, 2020 3:54 PM
To: Ryan Chen <[email protected]>
Cc: Brendan Higgins <[email protected]>; Benjamin Herrenschmidt <[email protected]>; Joel Stanley <[email protected]>; Andrew Jeffery <[email protected]>; linux-i2c at vger.kernel.org; openbmc at lists.ozlabs.org; linux-arm-kernel at lists.infradead.org; linux-aspeed at lists.ozlabs.org; linux-kernel at vger.kernel.org
Subject: Re: [PATCH v0 linux master] i2c/busses: Avoid i2c interrupt status clear race condition.
Quoted 8 lines · ryan_chen
In AST2600 there have a slow peripheral bus between CPU and i2c controller.
Therefore GIC i2c interrupt status clear have delay timing, when CPU issue write clear i2c controller interrupt status.
To avoid this issue, the driver need have read after write clear at i2c ISR.Signed-off-by ryan_chen <[email protected]>
v0? is it a prototype?
[Ryan Chen] It is not prototype it is official patch. And is there maybe a Fixes: tag for it?
[Ryan Chen] Yes it is a fix patch.
Quoted 32 lines
--- drivers/i2c/busses/i2c-aspeed.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)(quoted diff) +0
-0
b/drivers/i2c/busses/i2c-aspeed.cindex07c1993274c5..f51702d86a90100644drivers/i2c/busses/i2c-aspeed.c +4
-2@@ -603,6 +603,7 @@ static irqreturn_t aspeed_i2c_bus_irq(int irq, void *dev_id)/* Ack all interrupts except for Rx done */ writel (irq_received & ~ASPEED_I2CD_INTR_RX_DONE , bus->base +ASPEED_I2C_INTR_STS_REG ); +readl (bus->base +ASPEED_I2C_INTR_STS_REG ); irq_remaining = irq_received;#if IS_ENABLED (CONFIG_I2C_SLAVE )@@ -645,9 +646,11 @@ static irqreturn_t aspeed_i2c_bus_irq(int irq, void *dev_id) irq_received, irq_handled);/* Ack Rx done */ -if (irq_received &ASPEED_I2CD_INTR_RX_DONE ) +if (irq_received &ASPEED_I2CD_INTR_RX_DONE ) {writel (ASPEED_I2CD_INTR_RX_DONE , bus->base +ASPEED_I2C_INTR_STS_REG ); +readl (bus->base +ASPEED_I2C_INTR_STS_REG ); + }spin_unlock (&bus->lock);return irq_remaining ?IRQ_NONE :IRQ_HANDLED ; } --2.17.1
Quoted 3 lines
And is there maybe a Fixes: tag for it?
[Ryan Chen] Yes it is a fix patch.
I meant this (from submitting-patches.rst):
===
If your patch fixes a bug in a specific commit, e.g. you found an issue using ``git bisect``, please use the 'Fixes:' tag with the first 12 characters of the SHA-1 ID, and the one line summary. Do not split the tag across multiple lines, tags are exempt from the "wrap at 75 columns" rule in order to simplify parsing scripts. For example::
===
So, is it possible to identify a commit introducing the problem?
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://lists.ozlabs.org/pipermail/linux-aspeed/attachments/20200429/38bfa40a/attachment.sig>
Quoted 8 lines · ryan_chen
In AST2600 there have a slow peripheral bus between CPU
and i2c controller.
Therefore GIC i2c interrupt status clear have delay timing, when CPU issue write clear i2c controller interrupt status. To avoid this issue, the driver need have read after write
clear at i2c ISR.Signed-off-by ryan_chen <[email protected]>
--
Quoted 32 lines
--- drivers/i2c/busses/i2c-aspeed.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)drivers/i2c/busses/i2c-aspeed.c +1
-0@@ -603,6 +603,7 @@ static irqreturn_t aspeed_i2c_bus_irq(int irq, void *dev_id)/* Ack all interrupts except for Rx done */ writel(irq_received & ~ASPEED_I2CD_INTR_RX_DONE, bus->base + ASPEED_I2C_INTR_STS_REG); +readl (bus->base +ASPEED_I2C_INTR_STS_REG ); irq_remaining = irq_received;#if
IS_ENABLED(CONFIG_I2C_SLAVE)drivers/i2c/busses/i2c-aspeed.c +3
-1@@ -645,9 +646,11 @@ static irqreturn_t aspeed_i2c_bus_irq(int irq, void *dev_id) irq_received, irq_handled);/* Ack Rx done */ -if (irq_received &ASPEED_I2CD_INTR_RX_DONE ) +if (irq_received &ASPEED_I2CD_INTR_RX_DONE ) {writel (ASPEED_I2CD_INTR_RX_DONE , bus->base +ASPEED_I2C_INTR_STS_REG ); +readl (bus->base +ASPEED_I2C_INTR_STS_REG ); + }spin_unlock (&bus->lock);return irq_remaining ?IRQ_NONE :IRQ_HANDLED ;}
Benjamin Herrenschmidt
reviewerQuoted 4 lines · Wolfram Sang
Quoted 2 lines
And is there maybe a Fixes: tag for it?
[Ryan Chen] Yes it is a fix patch.I meant this (from
submitting-patches.rst):
It fixes the original implementation of the driver basically. It's just a classic posted-write fix. The write to clear the pending interrupt is asynchronous, so you can get spurrious ones if you return from the handler before it has percolated to the HW.
I assume it's just more visible on the 2600 because of the cores are significantly faster but the IO bus is still as dumb.
Ryan: You could always add a Fixed-by: tag that specifies the commit that added the initial driver...
Cheers,
Ben.
Wolfram Sang
AppliedQuoted 8 lines · ryan_chen
In AST2600 there have a slow peripheral bus between CPU
and i2c controller.
Therefore GIC i2c interrupt status clear have delay timing, when CPU issue write clear i2c controller interrupt status. To avoid this issue, the driver need have read after write
clear at i2c ISR.Signed-off-by ryan_chen <[email protected]>
Applied to for-current with a Fixes tag, thanks! Please, try to add one next time and please also check how the subsystem formats the $subject line.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://lists.ozlabs.org/pipermail/linux-aspeed/attachments/20200430/f27edc1e/attachment.sig>
Ryan Chen
maintainerQuoted 10 lines
In AST2600 there have a slow peripheral bus between CPU and i2c controller.
Therefore GIC i2c interrupt status clear have delay timing, when CPU issue write clear i2c controller interrupt status.
To avoid this issue, the driver need have read after write clear at i2c ISR.Signed-off-by ryan_chen <[email protected]>Applied to for-current with a Fixes tag, thanks! Please, try to add one next time and please also check how the subsystem formats the $subject line.
[Ryan Chen] Thanks your review, will add fixes tag at subject.
Ryan Chen
maintainerQuoted 10 lines
Quoted 2 lines
And is there maybe a Fixes: tag for it?
[Ryan Chen] Yes it is a fix patch.I meant this (from
submitting-patches.rst):It fixes the original implementation of the driver basically. It's just a classic posted-write fix. The write to clear the pending interrupt is asynchronous, so you can get spurrious ones if you return from the handler before it has percolated to the HW.
I assume it's just more visible on the 2600 because of the cores are significantly faster but the IO bus is still as dumb.
Ryan: You could always add a Fixed-by: tag that specifies the commit that added the initial driver...
[Ryan Chen] Thanks Ben.