On 32-bit (e.g. m68k):
fs/f2fs/gc.o: In function `f2fs_resize_fs':
gc.c:(.text+0x3056): undefined reference to `__umoddi3'
gc.c:(.text+0x30c4): undefined reference to `__udivdi3'
Fix this by using div_u64_rem() and div_u64() for 64-by-32 modulo resp.
division operations.
Reported-by: [email protected]
Fixes: d2ae7494d043bfaf ("f2fs: ioctl for removing a range from F2FS")
Signed-off-by: Geert Uytterhoeven <[email protected]>
---
This assumes BLKS_PER_SEC(sbi) is 32-bit.
#define BLKS_PER_SEC(sbi) \
((sbi)->segs_per_sec * (sbi)->blocks_per_seg)
Notes:
1. f2fs_sb_info.segs_per_sec and f2fs_sb_info.blocks_per_seg are both
unsigned int,
2. The multiplication is done in 32-bit arithmetic, hence the result
is of type unsigned int.
3. Is it guaranteed that the result will always fit in 32-bit, or can
this overflow?
4. fs/f2fs/debug.c:update_sit_info() assigns BLKS_PER_SEC(sbi) to
unsigned long long blks_per_sec, anticipating a 64-bit value.
---
fs/f2fs/gc.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
index 5b1076505ade9f84..c65f87f11de029f4 100644
--- a/fs/f2fs/gc.c
+++ b/fs/f2fs/gc.c
@@ -1438,13 +1438,15 @@ int f2fs_resize_fs(struct f2fs_sb_info *sbi, __u64 block_count)
unsigned int secs;
int gc_mode, gc_type;
int err = 0;
+ __u32 rem;
old_block_count = le64_to_cpu(F2FS_RAW_SUPER(sbi)->block_count);
if (block_count > old_block_count)
return -EINVAL;
/* new fs size should align to section size */
- if (block_count % BLKS_PER_SEC(sbi))
+ div_u64_rem(block_count, BLKS_PER_SEC(sbi), &rem);
+ if (rem)
return -EINVAL;
if (block_count == old_block_count)
@@ -1463,7 +1465,7 @@ int f2fs_resize_fs(struct f2fs_sb_info *sbi, __u64 block_count)
freeze_bdev(sbi->sb->s_bdev);
shrunk_blocks = old_block_count - block_count;
- secs = shrunk_blocks / BLKS_PER_SEC(sbi);
+ secs = div_u64(shrunk_blocks, BLKS_PER_SEC(sbi));
spin_lock(&sbi->stat_lock);
if (shrunk_blocks + valid_user_blocks(sbi) +
sbi->current_reserved_blocks + sbi->unusable_block_count +
--
2.17.1
Hi Chao,
On Fri, Jun 21, 2019 at 11:54 AM Chao Yu <[email protected]> wrote:
> Since the original patch hasn't been merged to upstream, I think we can merge
> this into original patch, how do you think?
Thanks, that's fine for me.
> On 2019/6/20 22:38, Geert Uytterhoeven wrote:
> > On 32-bit (e.g. m68k):
> >
> > fs/f2fs/gc.o: In function `f2fs_resize_fs':
> > gc.c:(.text+0x3056): undefined reference to `__umoddi3'
> > gc.c:(.text+0x30c4): undefined reference to `__udivdi3'
> >
> > Fix this by using div_u64_rem() and div_u64() for 64-by-32 modulo resp.
> > division operations.
> >
> > Reported-by: [email protected]
> > Fixes: d2ae7494d043bfaf ("f2fs: ioctl for removing a range from F2FS")
> > Signed-off-by: Geert Uytterhoeven <[email protected]>
> > ---
> > This assumes BLKS_PER_SEC(sbi) is 32-bit.
> >
> > #define BLKS_PER_SEC(sbi) \
> > ((sbi)->segs_per_sec * (sbi)->blocks_per_seg)
> >
> > Notes:
> > 1. f2fs_sb_info.segs_per_sec and f2fs_sb_info.blocks_per_seg are both
> > unsigned int,
> > 2. The multiplication is done in 32-bit arithmetic, hence the result
> > is of type unsigned int.
> > 3. Is it guaranteed that the result will always fit in 32-bit, or can
> > this overflow?
> > 4. fs/f2fs/debug.c:update_sit_info() assigns BLKS_PER_SEC(sbi) to
> > unsigned long long blks_per_sec, anticipating a 64-bit value.
> > ---
> > fs/f2fs/gc.c | 6 ++++--
> > 1 file changed, 4 insertions(+), 2 deletions(-)
> >
> > diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
> > index 5b1076505ade9f84..c65f87f11de029f4 100644
> > --- a/fs/f2fs/gc.c
> > +++ b/fs/f2fs/gc.c
> > @@ -1438,13 +1438,15 @@ int f2fs_resize_fs(struct f2fs_sb_info *sbi, __u64 block_count)
> > unsigned int secs;
> > int gc_mode, gc_type;
> > int err = 0;
> > + __u32 rem;
> >
> > old_block_count = le64_to_cpu(F2FS_RAW_SUPER(sbi)->block_count);
> > if (block_count > old_block_count)
> > return -EINVAL;
> >
> > /* new fs size should align to section size */
> > - if (block_count % BLKS_PER_SEC(sbi))
> > + div_u64_rem(block_count, BLKS_PER_SEC(sbi), &rem);
> > + if (rem)
> > return -EINVAL;
> >
> > if (block_count == old_block_count)
> > @@ -1463,7 +1465,7 @@ int f2fs_resize_fs(struct f2fs_sb_info *sbi, __u64 block_count)
> > freeze_bdev(sbi->sb->s_bdev);
> >
> > shrunk_blocks = old_block_count - block_count;
> > - secs = shrunk_blocks / BLKS_PER_SEC(sbi);
> > + secs = div_u64(shrunk_blocks, BLKS_PER_SEC(sbi));
> > spin_lock(&sbi->stat_lock);
> > if (shrunk_blocks + valid_user_blocks(sbi) +
> > sbi->current_reserved_blocks + sbi->unusable_block_count +
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
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
Hi Geert,
Since the original patch hasn't been merged to upstream, I think we can merge
this into original patch, how do you think?
On 2019/6/20 22:38, Geert Uytterhoeven wrote:
> On 32-bit (e.g. m68k):
>
> fs/f2fs/gc.o: In function `f2fs_resize_fs':
> gc.c:(.text+0x3056): undefined reference to `__umoddi3'
> gc.c:(.text+0x30c4): undefined reference to `__udivdi3'
>
> Fix this by using div_u64_rem() and div_u64() for 64-by-32 modulo resp.
> division operations.
>
> Reported-by: [email protected]
> Fixes: d2ae7494d043bfaf ("f2fs: ioctl for removing a range from F2FS")
> Signed-off-by: Geert Uytterhoeven <[email protected]>
> ---
> This assumes BLKS_PER_SEC(sbi) is 32-bit.
>
> #define BLKS_PER_SEC(sbi) \
> ((sbi)->segs_per_sec * (sbi)->blocks_per_seg)
>
> Notes:
> 1. f2fs_sb_info.segs_per_sec and f2fs_sb_info.blocks_per_seg are both
> unsigned int,
> 2. The multiplication is done in 32-bit arithmetic, hence the result
> is of type unsigned int.
> 3. Is it guaranteed that the result will always fit in 32-bit, or can
> this overflow?
> 4. fs/f2fs/debug.c:update_sit_info() assigns BLKS_PER_SEC(sbi) to
> unsigned long long blks_per_sec, anticipating a 64-bit value.
> ---
> fs/f2fs/gc.c | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
> index 5b1076505ade9f84..c65f87f11de029f4 100644
> --- a/fs/f2fs/gc.c
> +++ b/fs/f2fs/gc.c
> @@ -1438,13 +1438,15 @@ int f2fs_resize_fs(struct f2fs_sb_info *sbi, __u64 block_count)
> unsigned int secs;
> int gc_mode, gc_type;
> int err = 0;
> + __u32 rem;
>
> old_block_count = le64_to_cpu(F2FS_RAW_SUPER(sbi)->block_count);
> if (block_count > old_block_count)
> return -EINVAL;
>
> /* new fs size should align to section size */
> - if (block_count % BLKS_PER_SEC(sbi))
> + div_u64_rem(block_count, BLKS_PER_SEC(sbi), &rem);
> + if (rem)
> return -EINVAL;
>
> if (block_count == old_block_count)
> @@ -1463,7 +1465,7 @@ int f2fs_resize_fs(struct f2fs_sb_info *sbi, __u64 block_count)
> freeze_bdev(sbi->sb->s_bdev);
>
> shrunk_blocks = old_block_count - block_count;
> - secs = shrunk_blocks / BLKS_PER_SEC(sbi);
> + secs = div_u64(shrunk_blocks, BLKS_PER_SEC(sbi));
> spin_lock(&sbi->stat_lock);
> if (shrunk_blocks + valid_user_blocks(sbi) +
> sbi->current_reserved_blocks + sbi->unusable_block_count +
>
On 06/21, Geert Uytterhoeven wrote:
> Hi Chao,
>
> On Fri, Jun 21, 2019 at 11:54 AM Chao Yu <[email protected]> wrote:
> > Since the original patch hasn't been merged to upstream, I think we can merge
> > this into original patch, how do you think?
>
> Thanks, that's fine for me.
Merged the fix.
Thank you so much.
>
> > On 2019/6/20 22:38, Geert Uytterhoeven wrote:
> > > On 32-bit (e.g. m68k):
> > >
> > > fs/f2fs/gc.o: In function `f2fs_resize_fs':
> > > gc.c:(.text+0x3056): undefined reference to `__umoddi3'
> > > gc.c:(.text+0x30c4): undefined reference to `__udivdi3'
> > >
> > > Fix this by using div_u64_rem() and div_u64() for 64-by-32 modulo resp.
> > > division operations.
> > >
> > > Reported-by: [email protected]
> > > Fixes: d2ae7494d043bfaf ("f2fs: ioctl for removing a range from F2FS")
> > > Signed-off-by: Geert Uytterhoeven <[email protected]>
> > > ---
> > > This assumes BLKS_PER_SEC(sbi) is 32-bit.
> > >
> > > #define BLKS_PER_SEC(sbi) \
> > > ((sbi)->segs_per_sec * (sbi)->blocks_per_seg)
> > >
> > > Notes:
> > > 1. f2fs_sb_info.segs_per_sec and f2fs_sb_info.blocks_per_seg are both
> > > unsigned int,
> > > 2. The multiplication is done in 32-bit arithmetic, hence the result
> > > is of type unsigned int.
> > > 3. Is it guaranteed that the result will always fit in 32-bit, or can
> > > this overflow?
> > > 4. fs/f2fs/debug.c:update_sit_info() assigns BLKS_PER_SEC(sbi) to
> > > unsigned long long blks_per_sec, anticipating a 64-bit value.
> > > ---
> > > fs/f2fs/gc.c | 6 ++++--
> > > 1 file changed, 4 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
> > > index 5b1076505ade9f84..c65f87f11de029f4 100644
> > > --- a/fs/f2fs/gc.c
> > > +++ b/fs/f2fs/gc.c
> > > @@ -1438,13 +1438,15 @@ int f2fs_resize_fs(struct f2fs_sb_info *sbi, __u64 block_count)
> > > unsigned int secs;
> > > int gc_mode, gc_type;
> > > int err = 0;
> > > + __u32 rem;
> > >
> > > old_block_count = le64_to_cpu(F2FS_RAW_SUPER(sbi)->block_count);
> > > if (block_count > old_block_count)
> > > return -EINVAL;
> > >
> > > /* new fs size should align to section size */
> > > - if (block_count % BLKS_PER_SEC(sbi))
> > > + div_u64_rem(block_count, BLKS_PER_SEC(sbi), &rem);
> > > + if (rem)
> > > return -EINVAL;
> > >
> > > if (block_count == old_block_count)
> > > @@ -1463,7 +1465,7 @@ int f2fs_resize_fs(struct f2fs_sb_info *sbi, __u64 block_count)
> > > freeze_bdev(sbi->sb->s_bdev);
> > >
> > > shrunk_blocks = old_block_count - block_count;
> > > - secs = shrunk_blocks / BLKS_PER_SEC(sbi);
> > > + secs = div_u64(shrunk_blocks, BLKS_PER_SEC(sbi));
> > > spin_lock(&sbi->stat_lock);
> > > if (shrunk_blocks + valid_user_blocks(sbi) +
> > > sbi->current_reserved_blocks + sbi->unusable_block_count +
>
> Gr{oetje,eeting}s,
>
> Geert
>
> --
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
>
> 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