i2c: smbus: make i2c_smbus_read_block_data() safer
Subsystems (2)
PATCH RESEND 1/1 i2c: smbus: make i2c_smbus_read_block_data() safer
i2c_smbus_read_block_data() is dangerous to use because it may deliver up to I2C_SMBUS_BLOCK_MAX (32) bytes, which may be surprising to the caller. Callers tend to allocate buffers of sizes big enough to hold data from a well-behaving device and do not expect that i2c_smbus_read_block_data() may attempt to write more data than expected.
To make i2c_smbus_read_block_data() safer to use, change it so that it accepts size of the supplied buffer as another argument and ensure that it will not copy more data than the size of the buffer.
To allow users to gradually transition to the new API employ some macro trickery allowing calling i2c_smbus_read_block_data() with either 3 or 4 arguments. When called with 3 arguments it is assumed that the buffer size is I2C_SMBUS_BLOCK_MAX bytes. Once everyone is transitioned to the 4 argument form the macros should be removed.
---
This was sent out in October of '24 but unfortunately I did not get a reply from Wolfram...
Documentation/i2c/dev-interface.rst | 5 +++-- drivers/i2c/i2c-core-smbus.c | 21 +++++++++++++-------- include/linux/i2c.h | 19 +++++++++++++++++-- 3 files changed, 33 insertions(+), 12 deletions(-)
Documentation/i2c/dev-interface.rst +3 -2
@@ -161,9 +161,10 @@ for details) through the following functions:: __s32 i2c_smbus_process_call(int file, __u8 command, __u16 value); __s32 i2c_smbus_block_process_call(int file, __u8 command, __u8 length, __u8 *values); - __s32 i2c_smbus_read_block_data(int file, __u8 command, __u8 *values); + __s32 i2c_smbus_read_block_data(int file, __u8 command, __u8 length, + __u8 *values); __s32 i2c_smbus_write_block_data(int file, __u8 command, __u8 length, - __u8 *values); + const __u8 *values); All these transactions return -1 on failure; you can read errno to see what happened. The 'write' transactions return 0 on success; the
drivers/i2c/i2c-core-smbus.c +13 -8
@@ -208,11 +208,11 @@ s32 i2c_smbus_write_word_data(const struct i2c_client *client, u8 command,EXPORT_SYMBOL (i2c_smbus_write_word_data);/** -* i2c_smbus_read_block_data - SMBus "block read" protocol +* __i2c_smbus_read_block_data - SMBus "block read" protocol * @client: Handle to slave device * @command: Byte interpreted by slave -* @values: Byte array into which data will be read; big enough to hold -* the data returned by the slave. SMBus allows at most 32 bytes. +* @length: size of the @values array. SMBus allows at most 32 bytes +* @values: Byte array into which data will be read * * This executes the SMBus "block read" protocol, returning negative errno * else the number of data bytes in the slave's response.
@@ -222,22 +222,27 @@ EXPORT_SYMBOL(i2c_smbus_write_word_data);* support this; its emulation through I2C messaging relies on a specific * mechanism (I2C_M_RECV_LEN) which may not be implemented. */ -s32 i2c_smbus_read_block_data (const struct i2c_client *client,u8 command, -u8 *values) +s32 __i2c_smbus_read_block_data (const struct i2c_client *client,u8 command, +u8 length,u8 *values) {union i2c_smbus_data data; +int ret_len;int status; +if (length >I2C_SMBUS_BLOCK_MAX ) +return -EINVAL ; + status =i2c_smbus_xfer (client->adapter, client->addr, client->flags,I2C_SMBUS_READ , command,I2C_SMBUS_BLOCK_DATA , &data);if (status)return status; -memcpy (values, &data.block[1 ], data.block[0 ]); -return data.block[0 ]; + ret_len =min (length, data.block[0 ]); +memcpy (values, &data.block[1 ], ret_len); +return ret_len; } -EXPORT_SYMBOL (i2c_smbus_read_block_data); +EXPORT_SYMBOL (__i2c_smbus_read_block_data);/** * i2c_smbus_write_block_data - SMBus "block write" protocol
include/linux/i2c.h +17 -2
@@ -174,8 +174,23 @@ i2c_smbus_write_word_swapped(const struct i2c_client *client, }/* Returns the number of read bytes */ -s32 i2c_smbus_read_block_data (const struct i2c_client *client, -u8 command,u8 *values); +s32 __i2c_smbus_read_block_data (const struct i2c_client *client, +u8 command,u8 length,u8 *values); +/* +* This monstrosity allows to call i2c_smbus_read_block_data() with either +* 3 or 4 arguments and will be removed once all users have been switched +* to the 4 argument version. +*/ +#define __i2c_smbus_read_block_data_3arg (client, cmd, values) \ + __i2c_smbus_read_block_data(client, cmd, I2C_SMBUS_BLOCK_MAX, values) +#define __i2c_smbus_read_block_data_4arg (client, cmd, length, values) \ + __i2c_smbus_read_block_data(client, cmd, length, values) +#define __i2c_smbus_read_block_data_impl (_1, _2, _3, _4, impl, ...) impl +#define i2c_smbus_read_block_data(client, cmd, varargs...) \ + __i2c_smbus_read_block_data_impl(client, cmd, varargs, \ + __i2c_smbus_read_block_data_4arg, \ + __i2c_smbus_read_block_data_3arg) \ + (client, cmd, varargs)s32 i2c_smbus_write_block_data (const struct i2c_client *client,u8 command,u8 length,const u8 *values);/* Returns the number of read bytes */
Signature · 5 lines
2.55.0.795.g602f6c329a-goog -- Dmitry
Andi Shyti
maintainerHi Dmitry,
thanks for your patch. The core, indeed, does not know the size of the caller's buffer, so this can be, indeed, handled in the generic I2C code, instead of delegating it to individual host drivers.
...
Quoted 5 lines
Documentation/i2c/dev-interface.rst +0
-0
Documentation/i2c/dev-interface.rst for details) through the following functions:: · line 167
Quoted 9 lines
Documentation/i2c/dev-interface.rst +3
-2@@ -161,9 +161,10 @@ for details) through the following functions:: __s32 i2c_smbus_process_call(int file, __u8 command, __u16 value); __s32 i2c_smbus_block_process_call(int file, __u8 command, __u8 length, __u8 *values); - __s32 i2c_smbus_read_block_data(int file, __u8 command, __u8 *values); + __s32 i2c_smbus_read_block_data(int file, __u8 command, __u8 length, + __u8 *values); __s32 i2c_smbus_write_block_data(int file, __u8 command, __u8 length, - __u8 *values); + const __u8 *values);
The kernel and userspace APIs should remain aligned. If the kernel helper gains a buffer-length argument, it makes sense to consider the same improvement for libi2c.
This requires a coordinated i2c-tools change. Updating this prototype here alone would document a userspace function which does not yet exist.
Wolfram, do you have any opinion here?
Quoted 3 lines
Documentation/i2c/dev-interface.rst +0
-0All these transactions return -1 on failure; you can read errno to see what happened. The 'write' transactions return 0 on success; the
...
Documentation/i2c/dev-interface.rst
Quoted 12 lines
Documentation/i2c/dev-interface.rst +4
-2+ if (length > I2C_SMBUS_BLOCK_MAX) + return -EINVAL; + status = i2c_smbus_xfer(client->adapter, client->addr, client->flags, I2C_SMBUS_READ, command, I2C_SMBUS_BLOCK_DATA, &data); if (status) return status; - memcpy(values, &data.block[1], data.block[0]); - return data.block[0]; + ret_len = min(length, data.block[0]);
I don't like silently truncating the response.
If the device returns more data than the supplied buffer can hold, the caller should be notified. Otherwise, a truncated response is indistinguishable from a valid shorter response.
I think this should return an error, for example -EMSGSIZE, when data.block[0] is greater than length.
Thanks,
Andi
Quoted 3 lines
Documentation/i2c/dev-interface.rst +2
-0+ memcpy(values, &data.block[1], ret_len); + return ret_len; }
Hi Andi,
Quoted 28 lines · Andi Shyti
Hi Dmitry,
thanks for your patch. The core, indeed, does not know the size of the caller's buffer, so this can be, indeed, handled in the generic I2C code, instead of delegating it to individual host drivers.
...
Quoted 5 lines
Documentation/i2c/dev-interface.rst +0
-0Documentation/i2c/dev-interface.rst for details) through the following functions:: · line 167
Quoted 9 lines
Documentation/i2c/dev-interface.rst +3
-2@@ -161,9 +161,10 @@ for details) through the following functions:: __s32 i2c_smbus_process_call(int file, __u8 command, __u16 value); __s32 i2c_smbus_block_process_call(int file, __u8 command, __u8 length, __u8 *values); - __s32 i2c_smbus_read_block_data(int file, __u8 command, __u8 *values); + __s32 i2c_smbus_read_block_data(int file, __u8 command, __u8 length, + __u8 *values); __s32 i2c_smbus_write_block_data(int file, __u8 command, __u8 length, - __u8 *values); + const __u8 *values);The kernel and userspace APIs should remain aligned. If the kernel helper gains a buffer-length argument, it makes sense to consider the same improvement for libi2c.
Quoted 3 lines
This requires a coordinated i2c-tools change. Updating this prototype here alone would document a userspace function which does not yet exist.
Oops, I missed the fact that that document is for libi2c, not the kernel. I'll drop this chunk.
Quoted 24 lines
Wolfram, do you have any opinion here?
Quoted 3 lines
Documentation/i2c/dev-interface.rst +0
-0All these transactions return -1 on failure; you can read errno to see what happened. The 'write' transactions return 0 on success; the...
Documentation/i2c/dev-interface.rst
Quoted 12 lines
Documentation/i2c/dev-interface.rst +4
-2+ if (length > I2C_SMBUS_BLOCK_MAX) + return -EINVAL; + status = i2c_smbus_xfer(client->adapter, client->addr, client->flags, I2C_SMBUS_READ, command, I2C_SMBUS_BLOCK_DATA, &data); if (status) return status; - memcpy(values, &data.block[1], data.block[0]); - return data.block[0]; + ret_len = min(length, data.block[0]);I don't like silently truncating the response.
Quoted 6 lines
If the device returns more data than the supplied buffer can hold, the caller should be notified. Otherwise, a truncated response is indistinguishable from a valid shorter response.
I think this should return an error, for example
-EMSGSIZE, when data.block[0] is greater than length.
Yeah, I can do this.
Thanks.
Signature · 1 line
Dmitry