On 11.03.2025 10:46, Alejandro Vallejo wrote: > On Wed Mar 5, 2025 at 1:39 PM GMT, Jan Beulich wrote: >>> It's all quite perverse. Fortunately, looking at adjacent claims-related >>> code >>> xl seems to default to making a claim prior to populating the physmap and >>> cancelling the claim at the end of the meminit() hook so this is never a >>> real >>> problem. >>> >>> This tells me that the logic intent is to force early failure of >>> populate_physmap and nothing else. It's never active by the time ballooning >>> or >>> memory exchange matter at all. >> >> Ah yes, this I find more convincing. (Oddly enough this is all x86-only >> code.) > > Should I take this as an "ack" to the general plan of early returning on pages > <=0? I have a series pending that relies on it (the v2 of this[1]).
Not an ack, but an "I can accept this, but someone else will need to ack it". Jan > And would > rather defer its sending until this one get some form of nod. Otherwise I'll > integrate it in the other series so I can at least reduce remove dependencies > between things in-flight. > > Cheers, > Alejandro > > [1] > https://lore.kernel.org/xen-devel/20250304111000.9252-1-alejandro.vall...@cloud.com/
