i2c: microchip-core: fix "ghost" detections
Running i2c-detect currently produces an output akin to:
0 1 2 3 4 5 6 7 8 9 a b c d e f
00: 08 -- 0a -- 0c -- 0e --
10: 10 -- 12 -- 14 -- 16 -- UU 19 -- 1b -- 1d -- 1f
20: -- 21 -- 23 -- 25 -- 27 -- 29 -- 2b -- 2d -- 2f
30: -- -- -- -- -- -- -- -- 38 -- 3a -- 3c -- 3e --
40: 40 -- 42 -- 44 -- 46 -- 48 -- 4a -- 4c -- 4e --
50: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
60: 60 -- 62 -- 64 -- 66 -- 68 -- 6a -- 6c -- 6e --
70: 70 -- 72 -- 74 -- 76 --
This happens because for an i2c_msg with a len of 0 the driver will
mark the transmission of the message as a success once the START has
been sent, without waiting for the devices on the bus to respond with an
ACK/NAK. Since i2cdetect seems to run in a tight loop over all addresses
the NAK is treated as part of the next test for the next address.
Delete the fast path that marks a message as complete when idev->msg_len
is zero after sending a START/RESTART since this isn't a valid scenario.
CC: stable@vger.kernel.org
Fixes: 64a6f1c498
("i2c: add support for microchip fpga i2c controllers")
Signed-off-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Andi Shyti <andi.shyti@kernel.org>
Link: https://lore.kernel.org/r/20241218-outbid-encounter-b2e78b1cc707@spud
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
This commit is contained in:
parent
9a8f9320d6
commit
49e1f0fd0d
@ -287,8 +287,6 @@ static irqreturn_t mchp_corei2c_handle_isr(struct mchp_corei2c_dev *idev)
|
||||
ctrl &= ~CTRL_STA;
|
||||
writeb(idev->addr, idev->base + CORE_I2C_DATA);
|
||||
writeb(ctrl, idev->base + CORE_I2C_CTRL);
|
||||
if (idev->msg_len == 0)
|
||||
finished = true;
|
||||
break;
|
||||
case STATUS_M_ARB_LOST:
|
||||
idev->msg_err = -EAGAIN;
|
||||
|
Loading…
Reference in New Issue
Block a user