Re: [Qemu-devel] [PATCH] vpc: Remove useless parameters in get_sector_offset()

[Kevin Wolf]

Am 13.07.2017 um 14:42 hat Peter Maydell geschrieben:
> On 13 July 2017 at 13:37, Kevin Wolf <address@hidden> wrote:
> > All callers of get_sector_offset() pass write=false and err=NULL. Remove
> > the parameters.
> >
> > Signed-off-by: Kevin Wolf <address@hidden>
> > ---
> > Based on my block branch, specifically Peter's 'block/vpc.c: Handle write
> > failures in get_image_offset()'
> >
> >  block/vpc.c | 9 ++++-----
> >  1 file changed, 4 insertions(+), 5 deletions(-)
> >
> > diff --git a/block/vpc.c b/block/vpc.c
> > index bccc981..471f9ab 100644
> > --- a/block/vpc.c
> > +++ b/block/vpc.c
> > @@ -508,10 +508,9 @@ static inline int64_t 
> > get_image_offset(BlockDriverState *bs, uint64_t offset,
> >  }
> >
> >  static inline int64_t get_sector_offset(BlockDriverState *bs,
> > -                                        int64_t sector_num, bool write,
> > -                                        int *err)
> > +                                        int64_t sector_num)
> >  {
> > -    return get_image_offset(bs, sector_num * BDRV_SECTOR_SIZE, write, err);
> > +    return get_image_offset(bs, sector_num * BDRV_SECTOR_SIZE, false, 
> > NULL);
> >  }
> >
> >  /*
> > @@ -721,7 +720,7 @@ static int64_t coroutine_fn 
> > vpc_co_get_block_status(BlockDriverState *bs,
> >                 (sector_num << BDRV_SECTOR_BITS);
> >      }
> >
> > -    offset = get_sector_offset(bs, sector_num, false, NULL);
> > +    offset = get_sector_offset(bs, sector_num);
> >      start = offset;
> >      allocated = (offset != -1);
> >      *pnum = 0;
> > @@ -744,7 +743,7 @@ static int64_t coroutine_fn 
> > vpc_co_get_block_status(BlockDriverState *bs,
> >          if (nb_sectors == 0) {
> >              break;
> >          }
> > -        offset = get_sector_offset(bs, sector_num, false, NULL);
> > +        offset = get_sector_offset(bs, sector_num);
> >      } while (offset == -1);
> >
> >      return 0;
> 
> Just to copy my remark over from the other thread:
> this turns a function that could be used for writes into a function
> that doesn't look like it cares whether it's used for reads or
> writes but which will actually silently cause disk image
> corruption if you do ever use it in a write codepath. That's
> a bit of a bear trap as API design goes, which is why I opted
> to add the 'err' argument to get_image_offset() rather than
> just removing the 'write' argument.

Not sure if this is a problem in practice, but I'll leave it alone for
now. Once all of Eric's patches are in and make .bdrv_get_block_status
byte-based, the function should just go away anyway.

Kevin