mirror of https://github.com/torvalds/linux.git
49e1f0fd0d
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:
|
||
|---|---|---|
| .. | ||
| algos | ||
| busses | ||
| muxes | ||
| Kconfig | ||
| Makefile | ||
| i2c-atr.c | ||
| i2c-boardinfo.c | ||
| i2c-core-acpi.c | ||
| i2c-core-base.c | ||
| i2c-core-of-prober.c | ||
| i2c-core-of.c | ||
| i2c-core-slave.c | ||
| i2c-core-smbus.c | ||
| i2c-core.h | ||
| i2c-dev.c | ||
| i2c-mux.c | ||
| i2c-slave-eeprom.c | ||
| i2c-slave-testunit.c | ||
| i2c-smbus.c | ||
| i2c-stub.c | ||