> Date: Wed, 09 Jan 2019 17:25:12 +0900 (JST)
> From: YASUOKA Masahiko <yasu...@openbsd.org>
>
> On Wed, 09 Jan 2019 16:47:32 +0900 (JST)
> YASUOKA Masahiko <yasu...@openbsd.org> wrote:
> > The diff makes it can boot from paritions other than "a".
>
> Let me explain the diff.
>
> > Index: sys/arch/arm64/arm64/autoconf.c
> > ===================================================================
> > RCS file: /disk/cvs/openbsd/src/sys/arch/arm64/arm64/autoconf.c,v
> > retrieving revision 1.9
> > diff -u -p -r1.9 autoconf.c
> > --- sys/arch/arm64/arm64/autoconf.c 6 Feb 2018 20:35:21 -0000 1.9
> > +++ sys/arch/arm64/arm64/autoconf.c 4 Jan 2019 13:40:58 -0000
> > @@ -19,6 +19,7 @@
> > #include <sys/systm.h>
> > #include <sys/conf.h>
> > #include <sys/device.h>
> > +#include <sys/disklabel.h>
> > #include <sys/reboot.h>
> > #include <sys/socket.h>
> > #include <sys/hibernate.h>
> > @@ -68,7 +69,8 @@ diskconf(void)
> > {
> > size_t len;
> > char *p;
> > - dev_t tmpdev;
> > + dev_t tmpdev = NODEV;
> > + int part = 0;
> > extern uint8_t *bootmac;
> >
> > if (*boot_file != '\0')
> > @@ -82,6 +84,8 @@ diskconf(void)
> > else
> > len = strlen(boot_file);
> > bootdv = parsedisk(boot_file, len, 0, &tmpdev);
> > + if (tmpdev != NODEV)
> > + part = DISKPART(tmpdev);
> > }
> >
>
> parsedisk() parses the parition part. The diff is to keep it.
>
> > #if defined(NFSCLIENT)
> > @@ -105,7 +109,7 @@ diskconf(void)
> > else
> > printf("boot device: lookup %s failed \n", boot_file);
> >
> > - setroot(bootdv, 0, RB_USERREQ);
> > + setroot(bootdv, part, RB_USERREQ);
> > dumpconf();
> >
> > #ifdef HIBERNATE
>
> Previously 0 ("a" partition) is passed to setroot() always.
> By the diff, the partition kept above will be passed to setroot().
>
> Then the kernel will use the partition specified by the bootargs for
> the default root.
>
> > Index: sys/arch/arm64/stand/efiboot/efidev.c
> > ===================================================================
> > RCS file: /disk/cvs/openbsd/src/sys/arch/arm64/stand/efiboot/efidev.c,v
> > retrieving revision 1.2
> > diff -u -p -r1.2 efidev.c
> > --- sys/arch/arm64/stand/efiboot/efidev.c 25 Aug 2018 20:43:39 -0000
> > 1.2
> > +++ sys/arch/arm64/stand/efiboot/efidev.c 4 Jan 2019 13:40:58 -0000
> > @@ -484,7 +484,7 @@ efiopen(struct open_file *f, ...)
> > part = va_arg(ap, u_int);
> > va_end(ap);
> >
> > - if (unit != 0)
> > + if (unit != 0 || MAXPARTITIONS <= part)
> > return (ENXIO);
Can you make that "part >= MAXPARTITIONS". Easier to understand for
my brain!
> >
> > diskinfo.ed.blkio = disk;
> > @@ -512,7 +512,7 @@ efistrategy(void *devdata, int rw, daddr
> >
> > nsect = (size + DEV_BSIZE - 1) / DEV_BSIZE;
> > blk += DL_SECTOBLK(&dip->disklabel,
> > - dip->disklabel.d_partitions[B_PARTITION(dip->sc_part)].p_offset);
> > + dip->disklabel.d_partitions[dip->sc_part].p_offset);
> >
> > if (blk < 0)
> > error = EINVAL;
>
> Second chunk is to fix a misuse of B_PARTITION(). B_PARTITION()'s
> argument must be a "boot-style device number". dip->sc_part is usable
> just as it is. First chunk is to check the value more strict since
> it's used just as it is.
>
> > Index: sys/arch/arm64/stand/efiboot/exec.c
> > ===================================================================
> > RCS file: /disk/cvs/openbsd/src/sys/arch/arm64/stand/efiboot/exec.c,v
> > retrieving revision 1.5
> > diff -u -p -r1.5 exec.c
> > --- sys/arch/arm64/stand/efiboot/exec.c 8 Feb 2017 09:18:24 -0000
> > 1.5
> > +++ sys/arch/arm64/stand/efiboot/exec.c 4 Jan 2019 13:40:58 -0000
> > @@ -94,7 +94,7 @@ run_loadfile(u_long *marks, int howto)
> > char *cp;
> > void *fdt;
> >
> > - snprintf(args, sizeof(args) - 8, "%s:%s", cmd.bootdev, cmd.image);
> > + strlcpy(args, cmd.path, sizeof(args));
> > cp = args + strlen(args);
> >
> > *cp++ = ' ';
>
> cmd.bootdev is the "device" variable and cmd.image is the "image"
> variable of boot(8). They are used as defaults, but if a kernel is
> specified as an argument for "boot" command, they are not used.
> cmd.path is the real path that is used for loading a kernel. The diff
> is to use it as the bootargs as it is.
Thanks for the explanation. ok kettenis@ with the change suggested above.
