On 2012-11-01 17:13, Jordan Justen wrote: > On Wed, Oct 31, 2012 at 11:57 PM, Jan Kiszka <[email protected]> wrote: >> On 2012-11-01 03:55, Jordan Justen wrote: >>> On Tue, Oct 30, 2012 at 11:14 PM, Jan Kiszka <[email protected]> wrote: >>>> Hi Jordan, >>>> >>>> I was starring at pc_isa_bios_init and wondering why you are creating a >>>> copy of the system flash for the low ISA range instead of using an alias >>>> here as well, just like old_pc_system_rom_init does. That means the ISA >>>> BIOS range can run out of sync when the system flash is updated during >>>> runtime and requires a restart of QEMU then. Switching to an alias would >>>> also allow some code consolidation. Can you explain the idea behind the >>>> current version? >>> >>> I'm pretty sure I tried this and found that it did not work on the flash >>> device. >>> >>> I wrote an email to the list on Oct 17, 2011 about this, but I didn't >>> get a response. (Re: [Qemu-devel] [PATCH 2/4] pc: Support system flash >>> memory with pflash) >> >> Can you be more specific how/when it failed? A trivial test cannot >> confirm this so far. > > I reproduced this on 1b89fafe (where x86-flash was introduced). When I > setup the alias, I saw random data in the F000 segment. > > When I tried the same thing on the current master (patch attached), it > appeared to work correctly.
I vaguely remember issues with aliasing in the early memory region days, maybe that was causing it. As there is no conceptual reason to avoid an alias, let's file a refactoring patch. Jan
signature.asc
Description: OpenPGP digital signature
