On 06/03/25 at 02:11pm, Jiri Bohac wrote: > On Tue, Jun 03, 2025 at 07:02:06PM +0800, Baoquan He wrote: > > On 05/30/25 at 10:31pm, Jiri Bohac wrote: > > ......snip.. > > > @@ -582,7 +582,7 @@ static void __init arch_reserve_crashkernel(void) > > > > > > ret = parse_crashkernel(boot_command_line, memblock_phys_mem_size(), > > > &crash_size, &crash_base, > > > - &low_size, NULL, &high); > > > + &low_size, &cma_size, &high); > > > if (ret) > > > return; > > > > > > @@ -592,6 +592,7 @@ static void __init arch_reserve_crashkernel(void) > > > } > > > > > > reserve_crashkernel_generic(crash_size, crash_base, low_size, high); > > > + reserve_crashkernel_cma(cma_size); > > > > Wondering if ,high|low is still allowed (or needed) when ,cma is specified. > > Probably not needed but it works, totally independent of the > extra CMA-reserved area.
Allowing it can simplify the current code, while I can't imagine what cases need people to specify "crashkernel=xM,high crashkernel=xM,low crashkernel=zM,cma" at one time. Just personal thought, I haven't think of a strong reason to prevent it too. > > I saw no reason to artificially prevent it. > > -- > Jiri Bohac <[email protected]> > SUSE Labs, Prague, Czechia >
