Add tool to split Intel microcode into one file per Platform Id

[emaste]

Summary

Initial proof of concept tool to split per-Family-Model-Stepping microcode update files into one file per PlatformId

We currently do not look at the extended signature

[emaste]

Update model - family - stepping calculation

[sbruno]

tools/tools/ucode-split/ucode-split.c

54 ↗

(On Diff #42546)

Doesn't actually return char *. Reset to void ?

66 ↗

(On Diff #42546)

Needs #include <string.h> for sprintf().

[emaste]

tools/tools/ucode-split/ucode-split.c

54 ↗

(On Diff #42546)

Oops, I meant to have it return buf.

[markj]

tools/tools/ucode-split/ucode-split.c

32 ↗

(On Diff #42546)

Style: stdio.h isn't sorted correctly.

115 ↗

(On Diff #42546)

Is the intent here to create a single file per platform ID bit?

122 ↗

(On Diff #42546)

Style: extra space after sizeof.

130 ↗

(On Diff #42546)

Perhaps explicitly mask off the upper 24 bits?

141 ↗

(On Diff #42546)

resid might be a more idiomatic name.

155 ↗

(On Diff #42546)

Missing return.

[kib]

tools/tools/ucode-split/ucode-split.c

66 ↗

(On Diff #42546)

sprintf(3) is prototyped in stdio.h.

91 ↗

(On Diff #42546)

Style: blank line.

149 ↗

(On Diff #42546)

It is also reasonable to check that we write as much as we asked for. E.g. on fs full you get the truncated write.

155 ↗

(On Diff #42546)

No, main() is implicitly terminated by return (0);.

[emaste]

tools/tools/ucode-split/ucode-split.c

66 ↗

(On Diff #42546)

but is needed for strlen

115 ↗

(On Diff #42546)

Well, a single file per blob in one ucode file, which in practice is per PlatformId bit.

[emaste]

Review feedback, clean up warnings, and handle older-style microcode files with zero for data_size or total_size.

[emaste]

pass the signature to format_signature, no need for the full header

[emaste]

tools/tools/ucode-split/ucode-split.c

115 ↗

(On Diff #42546)

In some cases there's a unique blob per PlatformId bit:

06-05-00.01
06-05-00.02
06-05-00.08

In some cases there's only one blob that supports multiple PlatformIds:

06-55-03.97

In some cases there are multiple blobs that each support multiple PlatformIds:

06-17-0a.11
06-17-0a.44
06-17-0a.a0

[kib]

I think that I prefer to have this split as a mode of operations for cpucontrol(8).

Or add a loop into cpucontrol -u to try each blob from the file, instead of splitting into filesystem.

[emaste]

I think that I prefer to have this split as a mode of operations for cpucontrol(8).

That might be the eventual destination for this functionality. I put this up mainly for experimentation and so that @sbruno can try out things with the latest ucode drop.