[PATCH] misc: ibmvmc: fix buffer overflow in ibmvmc_recv_msg()

[PATCH] misc: ibmvmc: fix buffer overflow in ibmvmc_recv_msg()

[Junrui Luo]

The ibmvmc_recv_msg() function receives a msg_len value from the
hypervisor and uses it directly in h_copy_rdma() without validating.
This can result a buffer overflow.

Add a check to ensure msg_len does not exceed buffer->size before
calling h_copy_rdma().

Reported-by: Yuhao Jiang <danisjiang@gmail.com>
Reported-by: Junrui Luo <moonafterrain@outlook.com>
Fixes: 0eca353e7ae7 ("misc: IBM Virtual Management Channel Driver (VMC)")
Signed-off-by: Junrui Luo <moonafterrain@outlook.com>
---
 drivers/misc/ibmvmc.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/drivers/misc/ibmvmc.c b/drivers/misc/ibmvmc.c
index e5f935b5249d..e583a73a3592 100644
--- a/drivers/misc/ibmvmc.c
+++ b/drivers/misc/ibmvmc.c
@@ -1654,6 +1654,12 @@ static int ibmvmc_recv_msg(struct crq_server_adapter *adapter,
 	}
 
 	/* RDMA the data into the partition. */
+	if (msg_len > buffer->size) {
+		dev_err(adapter->dev, "Recv_msg: msg_len 0x%x exceeds buffer size 0x%x\n",
+			(unsigned int)msg_len, (unsigned int)buffer->size);
+		spin_unlock_irqrestore(&hmc->lock, flags);
+		return -EINVAL;
+	}
 	rc = h_copy_rdma(msg_len,
 			 adapter->riobn,
 			 buffer->dma_addr_remote,

---
base-commit: ea1013c1539270e372fc99854bc6e4d94eaeff66
change-id: 20251218-fixes-1fa33e72a87f

Best regards,
-- 
Junrui Luo <moonafterrain@outlook.com>

Re: [PATCH] misc: ibmvmc: fix buffer overflow in ibmvmc_recv_msg()

[Greg Kroah-Hartman]

On Thu, Dec 18, 2025 at 10:43:11PM +0800, Junrui Luo wrote:
> The ibmvmc_recv_msg() function receives a msg_len value from the
> hypervisor and uses it directly in h_copy_rdma() without validating.
> This can result a buffer overflow.
> 
> Add a check to ensure msg_len does not exceed buffer->size before
> calling h_copy_rdma().
> 
> Reported-by: Yuhao Jiang <danisjiang@gmail.com>
> Reported-by: Junrui Luo <moonafterrain@outlook.com>
> Fixes: 0eca353e7ae7 ("misc: IBM Virtual Management Channel Driver (VMC)")
> Signed-off-by: Junrui Luo <moonafterrain@outlook.com>
> ---
>  drivers/misc/ibmvmc.c | 6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/drivers/misc/ibmvmc.c b/drivers/misc/ibmvmc.c
> index e5f935b5249d..e583a73a3592 100644
> --- a/drivers/misc/ibmvmc.c
> +++ b/drivers/misc/ibmvmc.c
> @@ -1654,6 +1654,12 @@ static int ibmvmc_recv_msg(struct crq_server_adapter *adapter,
>  	}
>  
>  	/* RDMA the data into the partition. */
> +	if (msg_len > buffer->size) {
> +		dev_err(adapter->dev, "Recv_msg: msg_len 0x%x exceeds buffer size 0x%x\n",
> +			(unsigned int)msg_len, (unsigned int)buffer->size);

Don't allow broken hardware to spam the kernel log.  Make this a
debugging message instead.

And place the check above the comment?

thanks,

greg k-h