I tried to send the last message from my phone, and surprise it wasn't
formatted correctly, so it may have been marked as spam. repeating
myself again...
Ah, I guess I was wrong about logging OOM. I hadn’t hear about the
recommendations against it, but they make sense. Thanks for the
clarifications!
On Sat, Mar 30, 2019 at 10:27 AM Joe Perches <joe@xxxxxxxxxxx> wrote:
>
> On Sat, 2019-03-30 at 15:52 +0530, Rushikesh S Kadam wrote:
> > On Fri, Mar 29, 2019 at 04:30:18PM -0700, Nick Crews wrote:
> > > On Fri, Mar 29, 2019 at 1:03 PM Rushikesh S Kadam
> > > <rushikesh.s.kadam@xxxxxxxxx> wrote:
> > > > + ldr_xfer_ipc_frag = kzalloc(LOADER_SHIM_IPC_BUF_SIZE, GFP_KERNEL);
> > > > + if (!ldr_xfer_ipc_frag) {
> > > Log error here.
> > The error code is logged in calling function
> > load_fw_from_host(). Is that good enough?
> >
> > I believe the checkpatch script too, would
> > recommend against adding debug print for ENOMEM
> > error.
>
> The generic kernel allocation functions already do
> a dump_stack() on OOM conditions when called without
> __GFP_NOWARN so any additional OOM message isn't
> particularly useful.
>
> > Again, I thought it was against practise to log
> > "out of memory" debug prints in probe()
>
> Or anywhere else given the generic OOM stack dump.
>
> > But will add if you tell me this is the right way.
> >
> > > > + return -ENOMEM;
> > > > +
> > > > + loader_ishtp_cl = ishtp_cl_allocate(cl_device);
> > > > + if (!loader_ishtp_cl)
> > >
> > > log error here
>
> The ishtp_cl_allocate function just calls kmalloc then
> initializes the struct so an additional OOM message
> isn't useful here either.
>
>
