drivers/char/ipmi/ipmi_ipmb.c | 5 +++++ 1 file changed, 5 insertions(+)
IPMB doesn't use i2c reads, but the handler needs to set a value.
Otherwise an i2c read will return an uninitialised value from the bus
driver.
Fixes: 63c4eb347164 ("ipmi:ipmb: Add initial support for IPMI over IPMB")
Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
---
drivers/char/ipmi/ipmi_ipmb.c | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/drivers/char/ipmi/ipmi_ipmb.c b/drivers/char/ipmi/ipmi_ipmb.c
index 3a51e58b24875497699c7af7a4af1c73b47653f3..28818952a7a4bf814b95bdfb7046672ad4526cf2 100644
--- a/drivers/char/ipmi/ipmi_ipmb.c
+++ b/drivers/char/ipmi/ipmi_ipmb.c
@@ -202,11 +202,16 @@ static int ipmi_ipmb_slave_cb(struct i2c_client *client,
break;
case I2C_SLAVE_READ_REQUESTED:
+ *val = 0xff;
+ ipmi_ipmb_check_msg_done(iidev);
+ break;
+
case I2C_SLAVE_STOP:
ipmi_ipmb_check_msg_done(iidev);
break;
case I2C_SLAVE_READ_PROCESSED:
+ *val = 0xff;
break;
}
---
base-commit: b71e635feefc852405b14620a7fc58c4c80c0f73
change-id: 20260113-ipmb-read-init-5b97dfbcd3b9
Best regards,
--
Matt Johnston <matt@codeconstruct.com.au>
On Tue, Jan 13, 2026 at 05:41:34PM +0800, Matt Johnston wrote:
> IPMB doesn't use i2c reads, but the handler needs to set a value.
> Otherwise an i2c read will return an uninitialised value from the bus
> driver.
This is fine, I suppose. It's probably better to do this.
Are you actually using this code? How was it found?
-corey
>
> Fixes: 63c4eb347164 ("ipmi:ipmb: Add initial support for IPMI over IPMB")
> Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
> ---
> drivers/char/ipmi/ipmi_ipmb.c | 5 +++++
> 1 file changed, 5 insertions(+)
>
> diff --git a/drivers/char/ipmi/ipmi_ipmb.c b/drivers/char/ipmi/ipmi_ipmb.c
> index 3a51e58b24875497699c7af7a4af1c73b47653f3..28818952a7a4bf814b95bdfb7046672ad4526cf2 100644
> --- a/drivers/char/ipmi/ipmi_ipmb.c
> +++ b/drivers/char/ipmi/ipmi_ipmb.c
> @@ -202,11 +202,16 @@ static int ipmi_ipmb_slave_cb(struct i2c_client *client,
> break;
>
> case I2C_SLAVE_READ_REQUESTED:
> + *val = 0xff;
> + ipmi_ipmb_check_msg_done(iidev);
> + break;
> +
> case I2C_SLAVE_STOP:
> ipmi_ipmb_check_msg_done(iidev);
> break;
>
> case I2C_SLAVE_READ_PROCESSED:
> + *val = 0xff;
> break;
> }
>
>
> ---
> base-commit: b71e635feefc852405b14620a7fc58c4c80c0f73
> change-id: 20260113-ipmb-read-init-5b97dfbcd3b9
>
> Best regards,
> --
> Matt Johnston <matt@codeconstruct.com.au>
>
Hi Corey, On Tue, 2026-01-13 at 10:00 -0600, Corey Minyard wrote: > On Tue, Jan 13, 2026 at 05:41:34PM +0800, Matt Johnston wrote: > > IPMB doesn't use i2c reads, but the handler needs to set a value. > > Otherwise an i2c read will return an uninitialised value from the bus > > driver. > > This is fine, I suppose. It's probably better to do this. > > Are you actually using this code? How was it found? I'm not using IPMB, so haven't tested the patch on hardware. We received a patch for mctp-i2c read handling, that made me notice the uninitialised value problem there. I had a look over other i2c_slave_register() kernel users to see if the problem was common (mctp-i2c used ipmi-ipmb as a bit of a reference during development too). https://lore.kernel.org/netdev/20260108101829.1140448-1- zhangjian.3032@bytedance.com/ https://lore.kernel.org/netdev/20260113-mctp-read-fix-v1-1- 70c4b59c741c@codeconstruct.com.au/ Cheers, Matt
On Wed, Jan 14, 2026 at 10:10:34AM +0800, Matt Johnston wrote: > Hi Corey, > > On Tue, 2026-01-13 at 10:00 -0600, Corey Minyard wrote: > > On Tue, Jan 13, 2026 at 05:41:34PM +0800, Matt Johnston wrote: > > > IPMB doesn't use i2c reads, but the handler needs to set a value. > > > Otherwise an i2c read will return an uninitialised value from the bus > > > driver. > > > > This is fine, I suppose. It's probably better to do this. > > > > Are you actually using this code? How was it found? > > I'm not using IPMB, so haven't tested the patch on hardware. > We received a patch for mctp-i2c read handling, that made me notice > the uninitialised value problem there. I had a look over other > i2c_slave_register() kernel users to see if the problem was common > (mctp-i2c used ipmi-ipmb as a bit of a reference during development too). Ok, thanks for the history on this. It's in my tree. -corey > > https://lore.kernel.org/netdev/20260108101829.1140448-1- > zhangjian.3032@bytedance.com/ > https://lore.kernel.org/netdev/20260113-mctp-read-fix-v1-1- > 70c4b59c741c@codeconstruct.com.au/ > > Cheers, > Matt >
> IPMB doesn't use i2c reads, but the handler needs to set a value. > Otherwise an i2c read will return an uninitialised value from the bus > driver. See also once more: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/submitting-patches.rst?h=v6.19-rc5#n94 Regards, Markus
© 2016 - 2026 Red Hat, Inc.