uma: deduplicate uma_small_alloc

[bnovkov]

Summary

This patch refactors the UMA small alloc code and removes most UMA machine-dependent code.

The existing machine-dependent uma_small_alloc code was almost identical across all architectures, except for powerpc where using the direct map addresses involved extra steps in some cases.

The MI/MD split was replaced by a default uma_small_alloc implementation that can be overridden by architecture-specific code by defining the UMA_MD_SMALL_ALLOC symbol.
Furthermore, UMA_USE_DMAP was introduced to replace most UMA_MD_SMALL_ALLOC uses.

[kib]

sys/powerpc/powerpc/uma_machdep.c

41 ↗

(On Diff #138121)

Do you really need this guard? It seems to be always true.

[bnovkov]

sys/powerpc/powerpc/uma_machdep.c

41 ↗

(On Diff #138121)

I added it since powerpc's vmparam.h does not define it for 32 bit Book-E (lines 131-133).
However, I'm not that well acquainted with powerpc so I don't know if that combination is even possible.

[jhibbits]

sys/powerpc/powerpc/uma_machdep.c

41 ↗

(On Diff #138121)

Reading the code, it looks like we can unconditionally enable UMA_MD_SMALL_ALLOC for all powerpc. I think it wasn't set before is that we assumed a DMAP, which 32-bit Book-E cannot have. The code path has evolved to handle it, though.

[markj]

sys/vm/uma_core.c

2083

This can be static, right? Or does that make supporting UMA_MD_SMALL_ALLOC difficult?

2084

since you're moving this code anyway. The old u_int*_t typedefs are deprecated.

2168

[bnovkov]

Address multiple comments:

• enabled UMA_MD_SMALL_ALLOC unconditionally
• removed redundant include guard in ppc uma_machdep
• removed obsolete u_int8_t types

[bnovkov]

sys/vm/uma_core.c

2083

I don't think so, it would cause a 'static declaration follows non-static declaration' type of error on platforms that don't define UMA_MD_SMALL_ALLOC.

[markj]

sys/vm/uma_core.c

2083

Right, but you can guard the declarations with UMA_MD_SMALL_ALLOC. I don't have a strong feeling either way on that though.