On 3/3/22 19:45, Martin K. Petersen wrote:
On 3/1/22 21:35, Martin K. Petersen wrote:
+ if (min_xfer_bytes && opt_xfer_bytes & (min_xfer_bytes - 1)) {
+ sd_first_printk(KERN_WARNING, sdkp,
+ "Optimal transfer size %u bytes not a " \
+ "multiple of preferred minimum block " \
+ "size (%u bytes)\n",
+ opt_xfer_bytes, min_xfer_bytes);
+ return false;
+ }
Hmm ... what guarantees that min_xfer_bytes is a power of two? Did I
perhaps overlook something?
Nothing in the spec, obviously, but I think we'd have a lot of headaches
further up the stack if that wasn't the case.
Do you know of any devices that report something unusual here?
Hi Martin,
I'm not aware of any devices that report an unusual OPTIMAL TRANSFER LENGTH
GRANULARITY value.
Since this code is not in the hot path, how about using the modulo operator
instead of binary and?
Thanks,
Bart.
