On 7/26/23 14:07, Avri Altman wrote:
/**
@@ -2587,10 +2587,10 @@ static void ufshcd_prepare_req_desc_hdr(struct
ufshcd_lrb *lrbp, u8 *upiu_flags,
enum dma_data_direction cmd_dir, int ehs_length)
{
struct utp_transfer_req_desc *req_desc = lrbp->utr_descriptor_ptr;
+ struct request_desc_header *h = &req_desc->header;
u32 data_direction;
Maybe use the enum type as its strange to assign u32 to a nibble
Hi Avri,
I will make this change.
+ h->command_type = lrbp->command_type;
AFAIK the CT is always 1 in UFSHCI3.0?
That's also my understanding. I kept the existing logic since I think
that it's too early to drop UFSHCI 2 support?
+static void ufshcd_check_header_layout(void)
+{
+ BUILD_BUG_ON(((u8 *)&(struct request_desc_header){
+ .cci = 3})[0] != 3);
+
+ BUILD_BUG_ON(((u8 *)&(struct request_desc_header){
+ .ehs_length = 2})[1] != 2);
+
+ BUILD_BUG_ON(((u8 *)&(struct request_desc_header){
+ .enable_crypto = 1})[2]
+ != 0x80);
+
+ BUILD_BUG_ON((((u8 *)&(struct request_desc_header){
+ .command_type = 5,
+ .data_direction = 3,
+ .interrupt = 1,
+ })[3]) != ((5 << 4) | (3 << 1) | 1));
Isn't this checker assumes endianness hence requires the applicable #ifdef?
The above code builds fine on big endian and little endian systems
because of the #ifdefs that are present in the definition of struct
request_desc_header.
The script I use for compile testing of SCSI code is available here:
https://github.com/bvanassche/build-scsi-drivers/
Thanks,
Bart.
