Re: [PATCH v2 2/2] i2c: imx: fix locked bus on SMBus block-read of 0 (IRQ)

[Vincent Jardin <vjardin@free.fr>]

Hi bot,


> i2c: imx: fix locked bus on SMBus block-read of 0 (IRQ)
> [Severity: High]
> This is a pre-existing issue, but does this function also need to set the
> TXAK bit when len is exactly 1?
> 
> When the length byte is read here, the hardware immediately starts clocking
> in the next byte, which is the first data byte. If the block read length is
> exactly 1, this first data byte is also the last byte of the transfer.
> 
> Because TXAK is explicitly cleared earlier in i2c_imx_isr_read for length
> transfers, the controller will ACK this final data byte if TXAK is not set
> at this point.
> 
> If the slave holds SDA low as a result of receiving an ACK instead of a NACK
> on its final byte, would the subsequent attempt to generate a STOP condition
> in i2c_imx_isr_read_continue fail and permanently hang the bus?
> 
> The atomic path in i2c_imx_atomic_read appears to handle this safely by
> setting TXAK when reading the second-to-last byte.

once again, see the answers on v1:
  https://lore.kernel.org/all/ahRoKMLtrirbowbj@L30177.local/

regards,
  Vincent