translate-all: ensure host page mask is always extended with 1's
commit0c2d70c448b7853a91cfa63659aa3cc6630fb9be
authorPaolo Bonzini <pbonzini@redhat.com>
Wed, 2 Dec 2015 12:00:54 +0000 (2 13:00 +0100)
committerPaolo Bonzini <pbonzini@redhat.com>
Wed, 2 Dec 2015 12:12:30 +0000 (2 13:12 +0100)
tree7327dbaa1da19f0e397dfef405a45755ff61a92c
parent21a24302e85024dd7b2a151158adbc1f5dc5c4dd
translate-all: ensure host page mask is always extended with 1's

Anthony reported that >4GB guests on Xen with 32bit QEMU broke after
commit 4ed023c ("Round up RAMBlock sizes to host page sizes", 2015-11-05).

In that patch sizes are masked against qemu_host_page_size/mask which
are uintptr_t, and thus 32bit on a 32bit QEMU, even though the ram space
might be bigger than 4GB on Xen.

Since ram_addr_t is not available on user-mode emulation targets, ensure
that we get a sign extension when masking away the low bits of the address.
Remove the ~10 year old scary comment that the type of these variables
is probably wrong, with another equally scary comment.  The new comment
however does not have "???" in it, which is arguably an improvement.

For completeness use the alignment macros in linux-user and bsd-user
instead of manually doing an &.  linux-user and bsd-user are not affected
by the Xen issue, however.

Reviewed-by: Juan Quintela <quintela@redhat.com>
Reported-by: Anthony PERARD <anthony.perard@citrix.com>
Fixes: 4ed023ce2a39ab5812d33cf4d819def168965a7f
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
bsd-user/elfload.c
include/exec/cpu-all.h
linux-user/elfload.c
linux-user/mmap.c
translate-all.c
translate-common.c