Re: [dpdk-dev] [PATCH] bus/pci: set boot-up log prints to absolute minimum

[Thomas Monjalon <thomas@monjalon.net>]

26/04/2020 20:48, Jerin Jacob:
> On Sun, Apr 26, 2020 at 11:37 PM Thomas Monjalon <thomas@monjalon.net> wrote:
> > 26/04/2020 19:42, Jerin Jacob:
> > > On Sun, Apr 26, 2020 at 1:42 AM Thomas Monjalon <thomas@monjalon.net> wrote:
> > > > 06/02/2020 15:36, Jerin Jacob:
> > > > > On Thu, Feb 6, 2020 at 7:44 PM Thomas Monjalon <thomas@monjalon.net> wrote:
> > > > > > 21/01/2020 09:00, jerinj@marvell.com:
> > > > > > > From: Jerin Jacob <jerinj@marvell.com>
> > > > > > >
> > > > > > > Some machines may have a lot of PCI devices, logs from PCI probe
> > > > > > > creates a lot of clutter on boot-up, typically one needs
> > > > > > > to scroll the screen to find other issues in boot-up.
> > > > > > >
> > > > > > > This patch changes the loglevel of PCI probes to `debug`
> > > > > > > to reduce the clutter on default boot-up logs
> > > > > >
> > > > > > I think the PCI probe informations are... informational.
> > > > > > Maybe you are just not interested in info logs.
> > > > > > If this is the case, I suggest to change the log level at runtime.
> > > > >
> > > > > I am wondering, what would be the right balance, Following is DPDK
> > > > > startup output from octeontx2[1]
> > > > > It creates a lot of clutter in the "default" boot up. Why not enable
> > > > > below prints using log level at runtime?
> > > > > I believe it comes as a debug category, i.e information required to
> > > > > debug if something is not working,
> > > > > dpdk bind script already lists what is bound to DPDK.
> > > > >
> > > > > Suggestion to remove clutter?
> > > >
> > > > I suggest using dynamic log level in the PCI driver.
> > > > Unfortunately a lot of old DPDK code is still using the old log macros.
> > > > Some cleanup work is needed here.
> > >
> > > Sent an alternative fix to skip probing the devices with RTE_KDRV_NONE.
> >
> > No, a PCI PMD can work without a known kernel driver.
> > This is the case of mlx4/mlx5.
> 
> Yes. it can work with UNKNOWN, But It not with NONE.
> 
> > > I think that is the source of the problem.
> > > http://patches.dpdk.org/patch/69351/
> >
> > The source of the problem is just changing log levels dynamically
> > is not possible currently with PCI driver logs.
> 
> Assuming if we add, dynamic stuff what would the default log level for
> RTE_LOG(INFO, EAL, "PCI device "PCI_PRI_FMT" on NUMA socket %i\n",
> 
> If it is INFO, Still the problem persist in the default bootup. Right?

Yes

We need to know in INFO level which devices are successfully probed.
You want to decrease to DEBUG the level of logs announcing a probe.
I think it would be OK if adding an INFO log after probe success.