Good day,
On Tue, Nov 16, 2021 at 09:57:03PM +0800, Mark-PK Tsai wrote:
> > > > > Use %pe format string to print return error code which
> > > > > make the error message easier to understand.
> > > > >
> > > > > Signed-off-by: Mark-PK Tsai <mark-pk.tsai@xxxxxxxxxxxx>
> > > > > ---
> > > > > drivers/remoteproc/remoteproc_core.c | 4 ++--
> > > > > 1 file changed, 2 insertions(+), 2 deletions(-)
> > > > >
> > > > > diff --git a/drivers/remoteproc/remoteproc_core.c b/drivers/remoteproc/remoteproc_core.c
> > > > > index 502b6604b757..2242da320368 100644
> > > > > --- a/drivers/remoteproc/remoteproc_core.c
> > > > > +++ b/drivers/remoteproc/remoteproc_core.c
> > > > > @@ -575,8 +575,8 @@ static int rproc_handle_vdev(struct rproc *rproc, void *ptr,
> > > > > dma_get_mask(rproc->dev.parent));
> > > > > if (ret) {
> > > > > dev_warn(dev,
> > > > > - "Failed to set DMA mask %llx. Trying to continue... %x\n",
> > > > > - dma_get_mask(rproc->dev.parent), ret);
> > > > > + "Failed to set DMA mask %llx. Trying to continue... (%pe)\n",
> > > > > + dma_get_mask(rproc->dev.parent), ERR_PTR(ret));
> > > >
> > > > Macro ERR_PTR() is used to convert error codes to pointer type when
> > > > returning from a function - I fail to see how doing so in a dev_warn()
> > > > context can make the message easier to understand. Can you provide an
> > > > example?
> > >
> > > Hi,
> > >
> > > When dma_coerce_mask_and_coherent() fail, the output log will be as following.
> > >
> > > format log
> > > %x Trying to continue... fffffffb
> > > %d Trying to continue... -5
> > > %pe Trying to continue... -5 (if CONFIG_SYMBOLIC_ERRNAME is not set)
> > > %pe Trying to continue... -EIO (if CONFIG_SYMBOLIC_ERRNAME=y)
> >
> > When failing, functions dma_coerce_mask_and_coherent() returns -EIO.
> > Casting that to a (void *) with ERR_PTR() does not change that value.
> > Since variable @ret is already declared as "int" the real fix is to
> > change "%x" to "%d".
>
> There're some other drivers do the same thing in the recent kernel, so I think
> it's fine to casting the `ret` to a (void *) for more user friendly.
> But I suppose it would depend on the maintainer's opinion.
>
> So how about previous patch I sent, which also fix this issue by using %d.
>
> https://lore.kernel.org/lkml/20211102120805.27137-1-mark-pk.tsai@xxxxxxxxxxxx/
I had not seen Russell's reply in the link you posted above. I am not
completely sure about his solution but his insight and guidance in kernel
programming have precedence over mine.
I will take this patch.
Thanks,
Mathieu
>
