Re: [KJ] kmalloc to kzalloc patches for drivers/block [sane version]

[Jiri Slaby <jirislaby@gmail.com>]

Om Narasimhan wrote:
> Thanks for the comments.
> > >
> > > Signed off by Om Narasimhan <om.turyx@gmail.com>
> >
> > This is not the canonical format, per SubmittingPatches. It should be:
> >
> > Signed-off-by: Random J Developer <random@developer.example.org>
> OK. I would take care of it.
> > >  drivers/block/cciss.c    |    4 +--
> > >  drivers/block/cpqarray.c |   72 
> > +++++++++++++++-------------------------------
> > >  drivers/block/loop.c     |    4 +--
> > >  3 files changed, 25 insertions(+), 55 deletions(-)
> >
> > Your diffstat should have indicated to you that this should be split up
> > better. Please (re-)read SubmittingPatches. *One* logical change per
> > patch, most importantly.
> OK. I would resubmit.
> > >
> > > diff --git a/drivers/block/cciss.c b/drivers/block/cciss.c
> > > index 2cd3391..a800a69 100644
> > > --- a/drivers/block/cciss.c
> > > +++ b/drivers/block/cciss.c
> > > @@ -900,7 +900,7 @@ #if 0                             /* 'buf_size' 
> > member is 16-bits
> > >                               return -EINVAL;
> > >  #endif
> > >                       if (iocommand.buf_size > 0) {
> > > -                             buff = kmalloc(iocommand.buf_size, 
> > GFP_KERNEL);
> > > +                             buff = kzalloc(iocommand.buf_size, 
> > GFP_KERNEL);
> > >                               if (buff == NULL)
> > >                                       return -EFAULT;
> > >                       }
> > > @@ -911,8 +911,6 @@ #endif
> > >                                       kfree(buff);
> > >                                       return -EFAULT;
> > >                               }
> > > -                     } else {
> > > -                             memset(buff, 0, iocommand.buf_size);
> > >                       }
> > >                       if ((c = cmd_alloc(host, 0)) == NULL) {
> > >                               kfree(buff);
> >
> > This changes performance potentially, no? The memset before was
> > conditional upon (iocommand.Request.Type.Direction == XFER_WRITE) and
> > now the memory will always be zero'd.
> Yes, but not the functionality.
> if (iocommand.buf_size > 0), code allocates using kmalloc. if
> direction is XFER_WRITE, it does a copy_from_user(), and free()s the
> allocated buffer, not really caring what data came in from userspace.
> Else, it does memset(). So I could safely replace the kmalloc() with
> kzalloc() without compromising functionality.

Ok, this is something like I need 10 bytes of memory, so I request two memory 
pages for reserved use. It works, but it kills performance.

Why you zero memory that is not needed to be zeroed?

regards,
--
http://www.fi.muni.cz/~xslaby/            Jiri Slaby
faculty of informatics, masaryk university, brno, cz
e-mail: jirislaby gmail com, gpg pubkey fingerprint:
B674 9967 0407 CE62 ACC8  22A0 32CC 55C3 39D4 7A7E
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/