Re: [PATCH 1/1] scsi: ufs: remove ucd_rsp_dma_addr and ucd_prdt_dma_addr from ufshcd_lrb

[Bart Van Assche]

On 4/26/26 8:58 PM, ed.tsai@xxxxxxxxxxxx wrote:
> diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
> index 4805e40ed4d7..02fa61322e77 100644
> --- a/drivers/ufs/core/ufshcd.c
> +++ b/drivers/ufs/core/ufshcd.c
> @@ -621,7 +621,8 @@ static void ufshcd_print_tr(struct ufs_hba *hba, struct scsi_cmnd *cmd,
>   	ufshcd_hex_dump("UPIU REQ: ", lrbp->ucd_req_ptr,
>   			sizeof(struct utp_upiu_req));
>   	dev_err(hba->dev, "UPIU[%d] - Response UPIU phys@0x%llx\n", tag,
> -		(u64)lrbp->ucd_rsp_dma_addr);
> +		(u64)(lrbp->ucd_req_dma_addr +
> +		offsetof(struct utp_transfer_cmd_desc, response_upiu)));
>   	ufshcd_hex_dump("UPIU RSP: ", lrbp->ucd_rsp_ptr,
>   			sizeof(struct utp_upiu_rsp));
>   
> @@ -633,7 +634,8 @@ static void ufshcd_print_tr(struct ufs_hba *hba, struct scsi_cmnd *cmd,
>   	dev_err(hba->dev,
>   		"UPIU[%d] - PRDT - %d entries  phys@0x%llx\n",
>   		tag, prdt_length,
> -		(u64)lrbp->ucd_prdt_dma_addr);
> +		(u64)(lrbp->ucd_req_dma_addr +
> +		offsetof(struct utp_transfer_cmd_desc, prd_table)));

I don't think that it is useful to log DMA addresses and I prefer that 
this information would not be logged at all. Logging this information
might even involve a security risk. Here is some information about this
topic that comes from an LLM:
------------------------------------------------------------------------
Why is logging pointer addresses from kernel code considered a security 
risk?

Exposing kernel pointer addresses—a practice often referred to as 
pointer leaking—is considered a major security risk because it bypasses 
a fundamental defense mechanism called KASLR (Kernel Address Space 
Layout Randomization).
[ ... ]
2. Facilitating Exploit Chains

A pointer leak is rarely an exploit on its own, but it is almost always 
the first step in a sophisticated attack.

* Return-Oriented Programming (ROP): To hijack execution flow, an 
attacker needs "gadgets" (small snippets of existing code). Without 
knowing the exact addresses of these gadgets, their exploit will simply 
crash the system (a Denial of Service).

* Targeted Corruption: If an attacker wants to overwrite a specific 
security structure (like a process's UID to gain root access), they need 
the pointer to that specific object in kernel memory.
[ ... ]
------------------------------------------------------------------------

Thanks,

Bart.