Re: [PATCH 4/5] platform/x86/intel/ifs: Implement Array BIST test
From: Tony Luck
Date: Wed Feb 01 2023 - 14:22:30 EST
On Wed, Feb 01, 2023 at 07:19:15PM +0100, Greg KH wrote:
> It shouldn't be that hard, lots of people use them today.
>
> Try and see!
Extract from the first of our in-field-scan tests:
while (activate.start <= activate.stop) {
... trigger scan ...
if (status.chunk_num == activate.start) {
... check for too many retries on same chunk ...
} else {
activate.start = status.chunk_num;
}
}
using <linux/bitfield.h> becomes:
while (FIELD_GET(GENMASK(7, 0), activate) <= FIELD_GET(GENMASK(15, 8), activate) {
if (FIELD_GET(GENMASK(7, 0), status) == FIELD_GET(GENMASK(7, 0), activate) {
...
} else {
activate &= ~GENMASK(7, 0);
activate |= FIELD_PREP(GENMASK(7, 0), FIELD_GET(GENMASK(7, 0), status));
}
}
While I can make that more legible with some helper #defines for the
fields, it becomes more difficult to write, and no easier to read (since
I then have to chase down what the macros are doing).
If this were in some performance critical path, I might worry about
whether the generated code was good enough. But this code path isn't
remotely critical to anyone. The test takes up to 200 usec, so saving
a handful of cycles in the surrounding code will be in the noise.
-Tony