| Message ID | 20220705171350.1765454-2-andre.przywara@arm.com |
|---|---|
| State | Accepted |
| Delegated to: | Simon Glass |
| Headers | show |
| Series | fdt: introduce "apply-all" command | expand |
On Tue, 5 Jul 2022 at 11:14, Andre Przywara <andre.przywara@arm.com> wrote: > > The "fdt move" subcommand was using the provided DTB addresses directly, > without trying to "map" them into U-Boot's address space. This happened > to work since on the vast majority of "real" platforms there is a simple > 1:1 mapping of VA to PAs, so either value works fine. > > However this is not true on the sandbox, so the "fdt move" command fails > there miserably: > => fdt addr $fdtcontroladdr > => cp.l $fdtcontroladdr $fdt_addr_r 40 # simple memcpy works > => fdt move $fdtcontroladdr $fdt_addr_r > Segmentation fault > > Use the proper "map_sysmem" call to convert PAs to VAs, to make this > more robust in general and to enable operation in the sandbox. > > Signed-off-by: Andre Przywara <andre.przywara@arm.com> > --- > cmd/fdt.c | 6 +++--- > 1 file changed, 3 insertions(+), 3 deletions(-) Reviewed-by: Simon Glass <sjg@chromium.org>
On Tue, 5 Jul 2022 at 11:14, Andre Przywara <andre.przywara@arm.com> wrote: > > The "fdt move" subcommand was using the provided DTB addresses directly, > without trying to "map" them into U-Boot's address space. This happened > to work since on the vast majority of "real" platforms there is a simple > 1:1 mapping of VA to PAs, so either value works fine. > > However this is not true on the sandbox, so the "fdt move" command fails > there miserably: > => fdt addr $fdtcontroladdr > => cp.l $fdtcontroladdr $fdt_addr_r 40 # simple memcpy works > => fdt move $fdtcontroladdr $fdt_addr_r > Segmentation fault > > Use the proper "map_sysmem" call to convert PAs to VAs, to make this > more robust in general and to enable operation in the sandbox. > > Signed-off-by: Andre Przywara <andre.przywara@arm.com> > --- > cmd/fdt.c | 6 +++--- > 1 file changed, 3 insertions(+), 3 deletions(-) Reviewed-by: Simon Glass <sjg@chromium.org> Applied to u-boot-dm, thanks!
diff --git a/cmd/fdt.c b/cmd/fdt.c index 842e6cb634..abdc553b2b 100644 --- a/cmd/fdt.c +++ b/cmd/fdt.c @@ -211,11 +211,11 @@ static int do_fdt(struct cmd_tbl *cmdtp, int flag, int argc, char *const argv[]) /* * Set the address and length of the fdt. */ - working_fdt = (struct fdt_header *)hextoul(argv[2], NULL); + working_fdt = map_sysmem(hextoul(argv[2], NULL), 0); if (!fdt_valid(&working_fdt)) return 1; - newaddr = (struct fdt_header *)hextoul(argv[3], NULL); + newaddr = map_sysmem(hextoul(argv[3], NULL), 0); /* * If the user specifies a length, use that. Otherwise use the @@ -242,7 +242,7 @@ static int do_fdt(struct cmd_tbl *cmdtp, int flag, int argc, char *const argv[]) fdt_strerror(err)); return 1; } - set_working_fdt_addr((ulong)newaddr); + set_working_fdt_addr(map_to_sysmem(newaddr)); #ifdef CONFIG_OF_SYSTEM_SETUP /* Call the board-specific fixup routine */ } else if (strncmp(argv[1], "sys", 3) == 0) {
The "fdt move" subcommand was using the provided DTB addresses directly, without trying to "map" them into U-Boot's address space. This happened to work since on the vast majority of "real" platforms there is a simple 1:1 mapping of VA to PAs, so either value works fine. However this is not true on the sandbox, so the "fdt move" command fails there miserably: => fdt addr $fdtcontroladdr => cp.l $fdtcontroladdr $fdt_addr_r 40 # simple memcpy works => fdt move $fdtcontroladdr $fdt_addr_r Segmentation fault Use the proper "map_sysmem" call to convert PAs to VAs, to make this more robust in general and to enable operation in the sandbox. Signed-off-by: Andre Przywara <andre.przywara@arm.com> --- cmd/fdt.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)