Re: [PATCH 2/2] mtd: Check add_mtd_device() ret code

[Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>]

Hi Boris,

On Tue, Jan 29, 2019 at 12:03 PM Boris Brezillon <bbrezillon@xxxxxxxxxx> wrote:
> On Tue, 22 Jan 2019 13:31:59 +0100
> Boris Brezillon <bbrezillon@xxxxxxxxxx> wrote:
> > On Tue, 22 Jan 2019 12:21:11 +0100
> > Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
> > > On Wed, Jan 2, 2019 at 3:37 PM Boris Brezillon <bbrezillon@xxxxxxxxxx> wrote:
> > > > add_mtd_device() can fail. We should always check its return value
> > > > and gracefully handle the failure case. Fix the call sites where this
> > > > not done (in mtdpart.c) and add a __must_check attribute to the
> > > > prototype to avoid this kind of mistakes.
> > > >
> > > > Signed-off-by: Boris Brezillon <bbrezillon@xxxxxxxxxx>
> > > > ---
> > > > No Fixes or Cc-stable tag here, as this seems to have worked just fine
> > > > without checking add_mtd_device() ret code until we started to expose
> > > > MTD devices as NVMEM providers (queued for 4.21).
> > >
> > > Oh yes ;-)
> > >
> > > https://lore.kernel.org/lkml/87feac02-e955-1897-d4a4-d6d6d1082e45@xxxxxxxxx/t/
> > >
> > > Your patch is very similar to mine, so the crash is gone.
> >
> > Oops, sorry about that. I completely forgot about this patch. It seems
> > the discussion led to a different conclusion though (patch
> > allocate_partitions() to reject wrong parts early) and the v2 was never
> > sent (or I missed it). Anyway, I guess we should have done both (check
> > add_mtd_device() ret code everywhere and patch allocate_partitions() to
> > reject bad parts early).
> >
> > > However, the warning is still there:
> > >
> > >     m25p80 spi0.0: s25sl032p (4096 Kbytes)
> > >     3 fixed-partitions partitions found on MTD device spi0.0
> > >     Creating 3 MTD partitions on "spi0.0":
> > >     0x000000000000-0x000000080000 : "loader"
> > >     0x000000080000-0x000000600000 : "user"
> > >     mtd: partition "user" extends beyond the end of device "spi0.0" --
> > > size truncated to 0x380000
> > >     0x000000600000-0x000004000000 : "flash"
> > >     mtd: partition "flash" is out of reach -- disabled
> > >     ------------[ cut here ]------------
> > >     WARNING: CPU: 0 PID: 1 at drivers/mtd/mtdcore.c:571
> > > add_mtd_device+0x90/0x3b0
> > >
> > > Interestingly, only one partition is created, covering the full size of the
> > > device:
> > >
> > >     # cat /proc/partitions
> > >     major minor  #blocks  name
> > >
> > >       31        0       4096 mtdblock0
> > >
> > > While I would expect two partitions, "loader" and truncated "user":
> > >
> > >       31        0        512 mtdblock0
> > >       31        1       3584 mtdblock1
> >
> > Yes, makes sense, I guess your patch was better than mine :-/. Can you
> > try with the following diff applied and let me know if it solves the
> > problem?
>
> Gentle ping: is this diff fixing your problem, and do you want me to
> send a proper patch for it or should I let you send one?

Yes, the diff below fixes the partitions for me, so
Tested-by: Geert Uytterhoeven <geert+renesas@xxxxxxxxx>

Note that the warning is still there, but that's probably OK.

> > --- a/drivers/mtd/mtdpart.c
> > +++ b/drivers/mtd/mtdpart.c
> > @@ -724,16 +724,14 @@ int add_mtd_partitions(struct mtd_info *master,
> >  {
> >         struct mtd_part *slave;
> >         uint64_t cur_offset = 0;
> > -       int i, ret;
> > +       int i, ret, actual_nbparts = 0;
> >
> >         printk(KERN_NOTICE "Creating %d MTD partitions on \"%s\":\n", nbparts, master->name);
> >
> >         for (i = 0; i < nbparts; i++) {
> >                 slave = allocate_partition(master, parts + i, i, cur_offset);
> > -               if (IS_ERR(slave)) {
> > -                       ret = PTR_ERR(slave);
> > -                       goto err_del_partitions;
> > -               }
> > +               if (IS_ERR(slave))
> > +                       continue;
> >
> >                 mutex_lock(&mtd_partitions_mutex);
> >                 list_add(&slave->list, &mtd_partitions);
> > @@ -746,7 +744,7 @@ int add_mtd_partitions(struct mtd_info *master,
> >                         mutex_unlock(&mtd_partitions_mutex);
> >
> >                         free_partition(slave);
> > -                       goto err_del_partitions;
> > +                       continue;
> >                 }
> >
> >                 mtd_add_partition_attrs(slave);
> > @@ -754,14 +752,10 @@ int add_mtd_partitions(struct mtd_info *master,
> >                 parse_mtd_partitions(&slave->mtd, parts[i].types, NULL);
> >
> >                 cur_offset = slave->offset + slave->mtd.size;
> > +               actual_nbparts++;
> >         }
> >
> > -       return 0;
> > -
> > -err_del_partitions:
> > -       del_mtd_partitions(master);
> > -
> > -       return ret;
> > +       return actual_nbparts;
> >  }
> >
> >  static DEFINE_SPINLOCK(part_parser_lock);
> > @@ -1003,10 +997,10 @@ int parse_mtd_partitions(struct mtd_info *master, const char *const *types,
> >                 }
> >                 /* Found partitions! */
> >                 if (ret > 0) {
> > -                       err = add_mtd_partitions(master, pparts.parts,
> > +                       ret = add_mtd_partitions(master, pparts.parts,
> >                                                  pparts.nr_parts);
> >                         mtd_part_parser_cleanup(&pparts);
> > -                       return err ? err : pparts.nr_parts;
> > +                       return ret;
> >                 }
> >                 /*
> >                  * Stash the first error we see; only report it if no parser

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/