Skip to content

domain heap allocator changes

Jan Beulich xen-devel 1 version In review

Get series with b4
b4 am [email protected]

Versions newest first

  1. v1 Latest4 patches 8 repliesIn review

Patches in v1 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/4domain heap allocator changes - hide generic allocator routinesxen/common/page_alloc.c xen/include/xen/mm.h+88 -970 replies
2/4domain heap allocator changes - per-bit-width heap zonesxen/common/page_alloc.c+84 -470 replies
3/4domain heap allocator changes - remove bit width restrictionsxen/arch/x86/domain_build.c xen/common/memory.c xen/common/page_alloc.c +3 more+33 -248 replies
4/4domain heap allocator changes - per domain address restrictionxen/arch/x86/domain.c xen/arch/x86/domain_build.c xen/common/page_alloc.c +1 more+18 -20 replies

Reviews trailers per patch

No Reviewed-by, Acked-by or Tested-by trailers for v1 yet.

Read the v1 thread

Cover letter [PATCH 0/4] domain heap allocator changes

In order to (longer term) constrain 32-bit pv domains running on 64-bit hv to physical memory addresses it can actually control (namely use the m2p map on) without also limiting the amount (and range) of memory accessible to 64-bit domains, the following four patches will change the allocator so that a per-domain address size restriction can be established.

A positive side effect of this is that there no longer is a need to constrain DMA addresses to 30 bits (there's not going to be a DMA heap anymore), as by default memory will now be allocated from addresses above the DMA boundary (which is now purely for determining within what range memory should, if at all, be reserved, but not to artificially limit I/O-capable in what memory they are permitted to request), and could go away entirely if there wasn't the need to set aside some sufficiently low-address memory for the dom0 swiotlb), so setting the boundary to the more conventional 32 bits will still result in more constrained allocations to succeed (as long as there's memory meeting the constraint).

All patches have only been tested on x86.

Jan