| Message ID | 20191007231313.4700-4-jae.hyun.yoo@linux.intel.com |
|---|---|
| State | Changes Requested |
| Headers | show |
| Series | i2c: aspeed: Add buffer and DMA modes support | expand |
On Mon, Oct 07, 2019 at 04:13:11PM -0700, Jae Hyun Yoo wrote: > In case of master pending state, it should not trigger the master > command because this H/W is sharing the same data buffer for slave > and master operations, so this commit fixes the issue with making > the master command triggering happen when the state goes to active > state. nit: Makes sense, but can you explain what might happen without your change? > Signed-off-by: Jae Hyun Yoo <jae.hyun.yoo@linux.intel.com> Reviewed-by: Brendan Higgins <brendanhiggins@google.com> Thanks!
On 10/8/2019 1:31 PM, Brendan Higgins wrote: > On Mon, Oct 07, 2019 at 04:13:11PM -0700, Jae Hyun Yoo wrote: >> In case of master pending state, it should not trigger the master >> command because this H/W is sharing the same data buffer for slave >> and master operations, so this commit fixes the issue with making >> the master command triggering happen when the state goes to active >> state. > > nit: Makes sense, but can you explain what might happen without your > change? If we don't use this fix, a master command could corrupt data in the shared buffer if H/W is still on processing slave operation at the timing. >> Signed-off-by: Jae Hyun Yoo <jae.hyun.yoo@linux.intel.com> > > Reviewed-by: Brendan Higgins <brendanhiggins@google.com> Thanks a lot for your review! -Jae
On Tue, Oct 8, 2019 at 2:13 PM Jae Hyun Yoo <jae.hyun.yoo@linux.intel.com> wrote: > > On 10/8/2019 1:31 PM, Brendan Higgins wrote: > > On Mon, Oct 07, 2019 at 04:13:11PM -0700, Jae Hyun Yoo wrote: > >> In case of master pending state, it should not trigger the master > >> command because this H/W is sharing the same data buffer for slave > >> and master operations, so this commit fixes the issue with making > >> the master command triggering happen when the state goes to active > >> state. > > > > nit: Makes sense, but can you explain what might happen without your > > change? > > If we don't use this fix, a master command could corrupt data in the > shared buffer if H/W is still on processing slave operation at the > timing. Right, can you add that to the commit message? Is this trivially reproducible? We might want to submit this separately as a bugfix. Actually yeah, can you send this separately as a bugfix? I think we might want to include this in 5.4. Wolfram and Joel, what do you think?
On 10/7/19 4:13 PM, Jae Hyun Yoo wrote: > In case of master pending state, it should not trigger the master > command because this H/W is sharing the same data buffer for slave > and master operations, so this commit fixes the issue with making > the master command triggering happen when the state goes to active > state. > > Signed-off-by: Jae Hyun Yoo <jae.hyun.yoo@linux.intel.com> > --- > drivers/i2c/busses/i2c-aspeed.c | 9 +++++---- > 1 file changed, 5 insertions(+), 4 deletions(-) > > diff --git a/drivers/i2c/busses/i2c-aspeed.c b/drivers/i2c/busses/i2c-aspeed.c > index fa66951b05d0..40f6cf98d32e 100644 > --- a/drivers/i2c/busses/i2c-aspeed.c > +++ b/drivers/i2c/busses/i2c-aspeed.c > @@ -336,18 +336,19 @@ static void aspeed_i2c_do_start(struct aspeed_i2c_bus *bus) > struct i2c_msg *msg = &bus->msgs[bus->msgs_index]; > u8 slave_addr = i2c_8bit_addr_from_msg(msg); > > - bus->master_state = ASPEED_I2C_MASTER_START; > - > #if IS_ENABLED(CONFIG_I2C_SLAVE) > /* > * If it's requested in the middle of a slave session, set the master > * state to 'pending' then H/W will continue handling this master > * command when the bus comes back to the idle state. > */ > - if (bus->slave_state != ASPEED_I2C_SLAVE_INACTIVE) > + if (bus->slave_state != ASPEED_I2C_SLAVE_INACTIVE) { > bus->master_state = ASPEED_I2C_MASTER_PENDING; > + return; > + } > #endif /* CONFIG_I2C_SLAVE */ > > + bus->master_state = ASPEED_I2C_MASTER_START; > bus->buf_index = 0; > > if (msg->flags & I2C_M_RD) { > @@ -432,7 +433,7 @@ static u32 aspeed_i2c_master_irq(struct aspeed_i2c_bus *bus, u32 irq_status) > if (bus->slave_state != ASPEED_I2C_SLAVE_INACTIVE) > goto out_no_complete; > > - bus->master_state = ASPEED_I2C_MASTER_START; > + aspeed_i2c_do_start(bus); > } Shall we move the restart-master logic from master_irq to bus_irq? The reason being: master transaction cannot be restarted when aspeed-i2c is running in slave state and receives STOP interrupt, because aspeed_i2c_master_irq won't be called in this case. Cheers, Tao
Hi Tao, On 10/8/2019 3:00 PM, Tao Ren wrote: > On 10/7/19 4:13 PM, Jae Hyun Yoo wrote: >> In case of master pending state, it should not trigger the master >> command because this H/W is sharing the same data buffer for slave >> and master operations, so this commit fixes the issue with making >> the master command triggering happen when the state goes to active >> state. >> >> Signed-off-by: Jae Hyun Yoo <jae.hyun.yoo@linux.intel.com> >> --- >> drivers/i2c/busses/i2c-aspeed.c | 9 +++++---- >> 1 file changed, 5 insertions(+), 4 deletions(-) >> >> diff --git a/drivers/i2c/busses/i2c-aspeed.c b/drivers/i2c/busses/i2c-aspeed.c >> index fa66951b05d0..40f6cf98d32e 100644 >> --- a/drivers/i2c/busses/i2c-aspeed.c >> +++ b/drivers/i2c/busses/i2c-aspeed.c >> @@ -336,18 +336,19 @@ static void aspeed_i2c_do_start(struct aspeed_i2c_bus *bus) >> struct i2c_msg *msg = &bus->msgs[bus->msgs_index]; >> u8 slave_addr = i2c_8bit_addr_from_msg(msg); >> >> - bus->master_state = ASPEED_I2C_MASTER_START; >> - >> #if IS_ENABLED(CONFIG_I2C_SLAVE) >> /* >> * If it's requested in the middle of a slave session, set the master >> * state to 'pending' then H/W will continue handling this master >> * command when the bus comes back to the idle state. >> */ >> - if (bus->slave_state != ASPEED_I2C_SLAVE_INACTIVE) >> + if (bus->slave_state != ASPEED_I2C_SLAVE_INACTIVE) { >> bus->master_state = ASPEED_I2C_MASTER_PENDING; >> + return; >> + } >> #endif /* CONFIG_I2C_SLAVE */ >> >> + bus->master_state = ASPEED_I2C_MASTER_START; >> bus->buf_index = 0; >> >> if (msg->flags & I2C_M_RD) { >> @@ -432,7 +433,7 @@ static u32 aspeed_i2c_master_irq(struct aspeed_i2c_bus *bus, u32 irq_status) >> if (bus->slave_state != ASPEED_I2C_SLAVE_INACTIVE) >> goto out_no_complete; >> >> - bus->master_state = ASPEED_I2C_MASTER_START; >> + aspeed_i2c_do_start(bus); >> } > > Shall we move the restart-master logic from master_irq to bus_irq? The reason being: > master transaction cannot be restarted when aspeed-i2c is running in slave state and > receives STOP interrupt, because aspeed_i2c_master_irq won't be called in this case. Even in that case, master can be restarted properly because slave_irq will be called first because master is in MASTER_PENDING state, so the slave_irq handles the STOP interrupt as well, and then master_irq will be called with SLAVE_INACTIVE state so the aspeed_i2c_do_start can be called eventually. Also, this is right point to call the aspeed_i2c_do_start because master state will be changed to MASTER_START by the aspeed_i2c_do_start and we have to do remaining handling for the MASTER_START in the master_irq by falling through after the call. Thanks, Jae
On 10/8/2019 2:54 PM, Brendan Higgins wrote: > On Tue, Oct 8, 2019 at 2:13 PM Jae Hyun Yoo > <jae.hyun.yoo@linux.intel.com> wrote: >> >> On 10/8/2019 1:31 PM, Brendan Higgins wrote: >>> On Mon, Oct 07, 2019 at 04:13:11PM -0700, Jae Hyun Yoo wrote: >>>> In case of master pending state, it should not trigger the master >>>> command because this H/W is sharing the same data buffer for slave >>>> and master operations, so this commit fixes the issue with making >>>> the master command triggering happen when the state goes to active >>>> state. >>> >>> nit: Makes sense, but can you explain what might happen without your >>> change? >> >> If we don't use this fix, a master command could corrupt data in the >> shared buffer if H/W is still on processing slave operation at the >> timing. > > Right, can you add that to the commit message? Sure, will do that. > Is this trivially reproducible? We might want to submit this > separately as a bugfix. It's very rarely observed. > Actually yeah, can you send this separately as a bugfix? I think we > might want to include this in 5.4. Why not. I can send it separately but this patch series should wait for merging the bug fix to avoid context conflicts. > Wolfram and Joel, what do you think? >
On 10/8/19 3:45 PM, Jae Hyun Yoo wrote: > Hi Tao, > > On 10/8/2019 3:00 PM, Tao Ren wrote: >> On 10/7/19 4:13 PM, Jae Hyun Yoo wrote: >>> In case of master pending state, it should not trigger the master >>> command because this H/W is sharing the same data buffer for slave >>> and master operations, so this commit fixes the issue with making >>> the master command triggering happen when the state goes to active >>> state. >>> >>> Signed-off-by: Jae Hyun Yoo <jae.hyun.yoo@linux.intel.com> >>> --- >>> drivers/i2c/busses/i2c-aspeed.c | 9 +++++---- >>> 1 file changed, 5 insertions(+), 4 deletions(-) >>> >>> diff --git a/drivers/i2c/busses/i2c-aspeed.c b/drivers/i2c/busses/i2c-aspeed.c >>> index fa66951b05d0..40f6cf98d32e 100644 >>> --- a/drivers/i2c/busses/i2c-aspeed.c >>> +++ b/drivers/i2c/busses/i2c-aspeed.c >>> @@ -336,18 +336,19 @@ static void aspeed_i2c_do_start(struct aspeed_i2c_bus *bus) >>> struct i2c_msg *msg = &bus->msgs[bus->msgs_index]; >>> u8 slave_addr = i2c_8bit_addr_from_msg(msg); >>> - bus->master_state = ASPEED_I2C_MASTER_START; >>> - >>> #if IS_ENABLED(CONFIG_I2C_SLAVE) >>> /* >>> * If it's requested in the middle of a slave session, set the master >>> * state to 'pending' then H/W will continue handling this master >>> * command when the bus comes back to the idle state. >>> */ >>> - if (bus->slave_state != ASPEED_I2C_SLAVE_INACTIVE) >>> + if (bus->slave_state != ASPEED_I2C_SLAVE_INACTIVE) { >>> bus->master_state = ASPEED_I2C_MASTER_PENDING; >>> + return; >>> + } >>> #endif /* CONFIG_I2C_SLAVE */ >>> + bus->master_state = ASPEED_I2C_MASTER_START; >>> bus->buf_index = 0; >>> if (msg->flags & I2C_M_RD) { >>> @@ -432,7 +433,7 @@ static u32 aspeed_i2c_master_irq(struct aspeed_i2c_bus *bus, u32 irq_status) >>> if (bus->slave_state != ASPEED_I2C_SLAVE_INACTIVE) >>> goto out_no_complete; >>> - bus->master_state = ASPEED_I2C_MASTER_START; >>> + aspeed_i2c_do_start(bus); >>> } >> >> Shall we move the restart-master logic from master_irq to bus_irq? The reason being: >> master transaction cannot be restarted when aspeed-i2c is running in slave state and >> receives STOP interrupt, because aspeed_i2c_master_irq won't be called in this case. > > Even in that case, master can be restarted properly because slave_irq > will be called first because master is in MASTER_PENDING state, so the > slave_irq handles the STOP interrupt as well, and then master_irq will > be called with SLAVE_INACTIVE state so the aspeed_i2c_do_start can be > called eventually. I mean master_irq cannot be called when irq_remaining becomes 0 after slave_irq. Cheers, Tao
On 10/8/2019 4:15 PM, Tao Ren wrote: > On 10/8/19 3:45 PM, Jae Hyun Yoo wrote: >> Hi Tao, >> >> On 10/8/2019 3:00 PM, Tao Ren wrote: >>> On 10/7/19 4:13 PM, Jae Hyun Yoo wrote: >>>> In case of master pending state, it should not trigger the master >>>> command because this H/W is sharing the same data buffer for slave >>>> and master operations, so this commit fixes the issue with making >>>> the master command triggering happen when the state goes to active >>>> state. >>>> >>>> Signed-off-by: Jae Hyun Yoo <jae.hyun.yoo@linux.intel.com> >>>> --- >>>> drivers/i2c/busses/i2c-aspeed.c | 9 +++++---- >>>> 1 file changed, 5 insertions(+), 4 deletions(-) >>>> >>>> diff --git a/drivers/i2c/busses/i2c-aspeed.c b/drivers/i2c/busses/i2c-aspeed.c >>>> index fa66951b05d0..40f6cf98d32e 100644 >>>> --- a/drivers/i2c/busses/i2c-aspeed.c >>>> +++ b/drivers/i2c/busses/i2c-aspeed.c >>>> @@ -336,18 +336,19 @@ static void aspeed_i2c_do_start(struct aspeed_i2c_bus *bus) >>>> struct i2c_msg *msg = &bus->msgs[bus->msgs_index]; >>>> u8 slave_addr = i2c_8bit_addr_from_msg(msg); >>>> - bus->master_state = ASPEED_I2C_MASTER_START; >>>> - >>>> #if IS_ENABLED(CONFIG_I2C_SLAVE) >>>> /* >>>> * If it's requested in the middle of a slave session, set the master >>>> * state to 'pending' then H/W will continue handling this master >>>> * command when the bus comes back to the idle state. >>>> */ >>>> - if (bus->slave_state != ASPEED_I2C_SLAVE_INACTIVE) >>>> + if (bus->slave_state != ASPEED_I2C_SLAVE_INACTIVE) { >>>> bus->master_state = ASPEED_I2C_MASTER_PENDING; >>>> + return; >>>> + } >>>> #endif /* CONFIG_I2C_SLAVE */ >>>> + bus->master_state = ASPEED_I2C_MASTER_START; >>>> bus->buf_index = 0; >>>> if (msg->flags & I2C_M_RD) { >>>> @@ -432,7 +433,7 @@ static u32 aspeed_i2c_master_irq(struct aspeed_i2c_bus *bus, u32 irq_status) >>>> if (bus->slave_state != ASPEED_I2C_SLAVE_INACTIVE) >>>> goto out_no_complete; >>>> - bus->master_state = ASPEED_I2C_MASTER_START; >>>> + aspeed_i2c_do_start(bus); >>>> } >>> >>> Shall we move the restart-master logic from master_irq to bus_irq? The reason being: >>> master transaction cannot be restarted when aspeed-i2c is running in slave state and >>> receives STOP interrupt, because aspeed_i2c_master_irq won't be called in this case. >> >> Even in that case, master can be restarted properly because slave_irq >> will be called first because master is in MASTER_PENDING state, so the >> slave_irq handles the STOP interrupt as well, and then master_irq will >> be called with SLAVE_INACTIVE state so the aspeed_i2c_do_start can be >> called eventually. > > I mean master_irq cannot be called when irq_remaining becomes 0 after slave_irq. Ah, I see. It would be possibly happened. Probably we need to remove 'if (irq_remaining)' checking in bus_irq to make it call irqs always. Will fix the issue in the next round. Thanks, Jae
On Tue, 8 Oct 2019 at 21:54, Brendan Higgins <brendanhiggins@google.com> wrote: > > On Tue, Oct 8, 2019 at 2:13 PM Jae Hyun Yoo > <jae.hyun.yoo@linux.intel.com> wrote: > > > > On 10/8/2019 1:31 PM, Brendan Higgins wrote: > > > On Mon, Oct 07, 2019 at 04:13:11PM -0700, Jae Hyun Yoo wrote: > > >> In case of master pending state, it should not trigger the master > > >> command because this H/W is sharing the same data buffer for slave > > >> and master operations, so this commit fixes the issue with making > > >> the master command triggering happen when the state goes to active > > >> state. > > > > > > nit: Makes sense, but can you explain what might happen without your > > > change? > > > > If we don't use this fix, a master command could corrupt data in the > > shared buffer if H/W is still on processing slave operation at the > > timing. > > Right, can you add that to the commit message? > > Is this trivially reproducible? We might want to submit this > separately as a bugfix. > > Actually yeah, can you send this separately as a bugfix? I think we > might want to include this in 5.4. > > Wolfram and Joel, what do you think? Yes, good suggestion. A corruption fix should be merged I think. Always send bug fixes upstream with Fixes tags so they land in the stable tree. This is preferable to sending them separately to the openbmc for inclusion in that tree, and potentially reaches a wider audience. Cheers, Joel
diff --git a/drivers/i2c/busses/i2c-aspeed.c b/drivers/i2c/busses/i2c-aspeed.c index fa66951b05d0..40f6cf98d32e 100644 --- a/drivers/i2c/busses/i2c-aspeed.c +++ b/drivers/i2c/busses/i2c-aspeed.c @@ -336,18 +336,19 @@ static void aspeed_i2c_do_start(struct aspeed_i2c_bus *bus) struct i2c_msg *msg = &bus->msgs[bus->msgs_index]; u8 slave_addr = i2c_8bit_addr_from_msg(msg); - bus->master_state = ASPEED_I2C_MASTER_START; - #if IS_ENABLED(CONFIG_I2C_SLAVE) /* * If it's requested in the middle of a slave session, set the master * state to 'pending' then H/W will continue handling this master * command when the bus comes back to the idle state. */ - if (bus->slave_state != ASPEED_I2C_SLAVE_INACTIVE) + if (bus->slave_state != ASPEED_I2C_SLAVE_INACTIVE) { bus->master_state = ASPEED_I2C_MASTER_PENDING; + return; + } #endif /* CONFIG_I2C_SLAVE */ + bus->master_state = ASPEED_I2C_MASTER_START; bus->buf_index = 0; if (msg->flags & I2C_M_RD) { @@ -432,7 +433,7 @@ static u32 aspeed_i2c_master_irq(struct aspeed_i2c_bus *bus, u32 irq_status) if (bus->slave_state != ASPEED_I2C_SLAVE_INACTIVE) goto out_no_complete; - bus->master_state = ASPEED_I2C_MASTER_START; + aspeed_i2c_do_start(bus); } #endif /* CONFIG_I2C_SLAVE */
In case of master pending state, it should not trigger the master command because this H/W is sharing the same data buffer for slave and master operations, so this commit fixes the issue with making the master command triggering happen when the state goes to active state. Signed-off-by: Jae Hyun Yoo <jae.hyun.yoo@linux.intel.com> --- drivers/i2c/busses/i2c-aspeed.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-)