| Message ID | 20200908224006.25636-22-digetx@gmail.com |
|---|---|
| State | Changes Requested |
| Headers | show |
| Series | Improvements for Tegra I2C driver | expand |
| Context | Check | Description |
|---|---|---|
| tagr/GVS | success | 1124251 |
| tagr/GVS-1122491 | pending | None |
| tagr/GVS-1122491 | fail | None |
| tagr/GVS | pending | 1122533 |
| tagr/GVS | success | 1122533 |
| tagr/GVS | pending | 1124251 |
On Wed, Sep 09, 2020 at 01:39:53AM +0300, Dmitry Osipenko wrote: > The DMA code path has been tested well enough and the DMA configuration > performed by tegra_i2c_config_fifo_trig() shouldn't ever fail in practice. > Hence let's remove the obscure transfer-mode switching in order to have a > cleaner and simpler code. Now I2C transfer will be failed if DMA > configuration fails. > > Signed-off-by: Dmitry Osipenko <digetx@gmail.com> > --- > drivers/i2c/busses/i2c-tegra.c | 18 ++++++++++-------- > 1 file changed, 10 insertions(+), 8 deletions(-) I'm not sure that's a good idea. It's always possible that the DMA setup is going to break because of something that's not related to the I2C driver itself. Having the system completely break instead of falling back to PIO mode seems like it would only complicate troubleshooting any such issues. Thierry
17.09.2020 14:47, Thierry Reding пишет: > On Wed, Sep 09, 2020 at 01:39:53AM +0300, Dmitry Osipenko wrote: >> The DMA code path has been tested well enough and the DMA configuration >> performed by tegra_i2c_config_fifo_trig() shouldn't ever fail in practice. >> Hence let's remove the obscure transfer-mode switching in order to have a >> cleaner and simpler code. Now I2C transfer will be failed if DMA >> configuration fails. >> >> Signed-off-by: Dmitry Osipenko <digetx@gmail.com> >> --- >> drivers/i2c/busses/i2c-tegra.c | 18 ++++++++++-------- >> 1 file changed, 10 insertions(+), 8 deletions(-) > > I'm not sure that's a good idea. It's always possible that the DMA setup > is going to break because of something that's not related to the I2C > driver itself. Having the system completely break instead of falling > back to PIO mode seems like it would only complicate troubleshooting any > such issues. That code has zero test coverage because this problem never happens in practice, hence it should be better to have it removed. We may consider re-adding it back if there will be a real-world incident, okay?
On Wed, 09 Sep 2020 01:39:53 +0300, Dmitry Osipenko wrote: > The DMA code path has been tested well enough and the DMA configuration > performed by tegra_i2c_config_fifo_trig() shouldn't ever fail in practice. > Hence let's remove the obscure transfer-mode switching in order to have a > cleaner and simpler code. Now I2C transfer will be failed if DMA > configuration fails. > > Signed-off-by: Dmitry Osipenko <digetx@gmail.com> > --- > drivers/i2c/busses/i2c-tegra.c | 18 ++++++++++-------- > 1 file changed, 10 insertions(+), 8 deletions(-) Tested-by: Thierry Reding <treding@nvidia.com>
On Thu, Sep 17, 2020 at 06:03:04PM +0300, Dmitry Osipenko wrote: > 17.09.2020 14:47, Thierry Reding пишет: > > On Wed, Sep 09, 2020 at 01:39:53AM +0300, Dmitry Osipenko wrote: > >> The DMA code path has been tested well enough and the DMA configuration > >> performed by tegra_i2c_config_fifo_trig() shouldn't ever fail in practice. > >> Hence let's remove the obscure transfer-mode switching in order to have a > >> cleaner and simpler code. Now I2C transfer will be failed if DMA > >> configuration fails. > >> > >> Signed-off-by: Dmitry Osipenko <digetx@gmail.com> > >> --- > >> drivers/i2c/busses/i2c-tegra.c | 18 ++++++++++-------- > >> 1 file changed, 10 insertions(+), 8 deletions(-) > > > > I'm not sure that's a good idea. It's always possible that the DMA setup > > is going to break because of something that's not related to the I2C > > driver itself. Having the system completely break instead of falling > > back to PIO mode seems like it would only complicate troubleshooting any > > such issues. > > That code has zero test coverage because this problem never happens in > practice, hence it should be better to have it removed. We may consider > re-adding it back if there will be a real-world incident, okay? Again, I think throwing out fallbacks and error messages out the window just because they "don't happen in practice" is misguided. Just because they don't *usually* happen doesn't mean they can't happen. And in case they do happen we absolutely do want some way of dealing with it rather than just have the driver stop working without any explanation. Thierry
diff --git a/drivers/i2c/busses/i2c-tegra.c b/drivers/i2c/busses/i2c-tegra.c index 76d3a05b10fc..574b59a8b2a2 100644 --- a/drivers/i2c/busses/i2c-tegra.c +++ b/drivers/i2c/busses/i2c-tegra.c @@ -940,8 +940,7 @@ static irqreturn_t tegra_i2c_isr(int irq, void *dev_id) return IRQ_HANDLED; } -static void tegra_i2c_config_fifo_trig(struct tegra_i2c_dev *i2c_dev, - size_t len) +static int tegra_i2c_config_fifo_trig(struct tegra_i2c_dev *i2c_dev, size_t len) { u32 val, reg; u8 dma_burst; @@ -992,12 +991,10 @@ static void tegra_i2c_config_fifo_trig(struct tegra_i2c_dev *i2c_dev, if (ret < 0) { dev_err(i2c_dev->dev, "DMA slave config failed: %d\n", ret); - dev_err(i2c_dev->dev, "falling back to PIO\n"); - tegra_i2c_release_dma(i2c_dev); - i2c_dev->is_curr_dma_xfer = false; - } else { - goto out; + return ret; } + + goto out; } if (i2c_dev->hw->has_mst_fifo) @@ -1008,6 +1005,8 @@ static void tegra_i2c_config_fifo_trig(struct tegra_i2c_dev *i2c_dev, I2C_FIFO_CONTROL_RX_TRIG(1); out: i2c_writel(i2c_dev, val, reg); + + return 0; } static unsigned long @@ -1141,7 +1140,10 @@ static int tegra_i2c_xfer_msg(struct tegra_i2c_dev *i2c_dev, i2c_dev->is_curr_dma_xfer = (xfer_size > I2C_PIO_MODE_PREFERRED_LEN) && i2c_dev->dma_buf && !i2c_dev->is_curr_atomic_xfer; - tegra_i2c_config_fifo_trig(i2c_dev, xfer_size); + + err = tegra_i2c_config_fifo_trig(i2c_dev, xfer_size); + if (err) + return err; /* * Transfer time in mSec = Total bits / transfer rate
The DMA code path has been tested well enough and the DMA configuration performed by tegra_i2c_config_fifo_trig() shouldn't ever fail in practice. Hence let's remove the obscure transfer-mode switching in order to have a cleaner and simpler code. Now I2C transfer will be failed if DMA configuration fails. Signed-off-by: Dmitry Osipenko <digetx@gmail.com> --- drivers/i2c/busses/i2c-tegra.c | 18 ++++++++++-------- 1 file changed, 10 insertions(+), 8 deletions(-)