[LTP] [PATCH v3 1/7] lib: Add library functions for sync related syscalls

[Sumit Garg]

On Wed, 20 Feb 2019 at 17:34, Cyril Hrubis <chrubis@suse.cz> wrote:
>
> Hi!
> > > Maybe we just need a tst_dev_bytes_written() function that would return
> > > the bytes written since the last call of the function so that we could
> > > do:
> > >
> > >         fd = SAFE_OPEN(...);
> > >
> > >         tst_dev_blocks_written(tst_device.dev);
> > >
> > >         tst_fill_fd(fd, 0, TST_MB, size_mb);
> > >
> > >         TEST(fdsync(fd));
> > >
> > >         if (TST_RET)
> > >                 tst_brk(TFAIL | TTERRNO, "syncfd() failed with %i", TST_RET);
> > >
> > >         written = tst_dev_blocks_written(tst_device.dev);
> > >
> > >         SAFE_CLOSE(fd);
> > >
> > >         if (written >= size_mb * TST_DEV_BLOCKS_IN_MB)
> > >                 tst_res(TPASS, ...);
> > >         else
> > >                 tst_res(TFAIL, ...);
> > >
> >
> > Seems to be nicer approach. So rather than new file should I add
> > tst_dev_bytes_written() function in "lib/tst_device.c" file?
>
> Sounds good to me.
>
> > > The test a bit longer, but the library functions are more reusable, if
> > > you do 'git grep -B 1 SAFE_WRITE' you can see that the tst_fill_fd
> > > function that loops over SAFE_WRITE could be used in a few places
> > > already.
> > >
> >
> > I am not able to locate tst_fill_fd function. Are you referring to
> > tst_fill_file function? If yes, then we could split that function to
> > create tst_fill_fd function.
>
> There is none, I was trying to state that there are several places that
> use the same pattern and that such function would be generally useful.
>

Ah, ok.

> And yes, buiding tst_fill_file on the top of tst_fill_fd sounds like a
> good approach.
>

Will incorporate in v4.

-Sumit

> --
> Cyril Hrubis
> chrubis@suse.cz