no printf in drivers code

no printf in drivers code

[Thomas Monjalon]

Hello,

We have too many drivers using printf() instead of rte_log().
Please propose a plan to replace or remove the calls to printf().
If no plan, I could propose one but you may not like it :)

Affected drivers are:
	baseband/acc
	bus/dpaa
	bus/fslmc
	crypto/caam_jr
	crypto/ccp
	dma/ioat
	net/dpaa
	net/dpaa2
	net/nfp
	net/qede
	net/txgbe
	raw/ifpga

PS: printf is allowed for tests and dumps.

RE: no printf in drivers code

[Honnappa Nagarahalli]

> -----Original Message-----
> From: Thomas Monjalon <thomas@monjalon.net>
> Sent: Friday, February 3, 2023 7:57 AM
> To: Nicolas Chautru <nicolas.chautru@intel.com>; hemant.agrawal@nxp.com;
> Sachin Saxena <sachin.saxena@oss.nxp.com>; Gagandeep Singh
> <g.singh@nxp.com>; Sunil Uttarwar <sunilprakashrao.uttarwar@amd.com>;
> Bruce Richardson <bruce.richardson@intel.com>; Conor Walsh
> <conor.walsh@intel.com>; Chaoyong He <chaoyong.he@corigine.com>; Niklas
> Soderlund <niklas.soderlund@corigine.com>; Rasesh Mody
> <rmody@marvell.com>; Devendra Singh Rawat <dsinghrawat@marvell.com>;
> Jiawen Wu <jiawenwu@trustnetic.com>; Jian Wang
> <jianwang@trustnetic.com>; Rosen Xu <rosen.xu@intel.com>; Tianfei zhang
> <tianfei.zhang@intel.com>
> Cc: dev@dpdk.org; Ferruh Yigit <ferruh.yigit@amd.com>; Andrew Rybchenko
> <andrew.rybchenko@oktetlabs.ru>; david.marchand@redhat.com
> Subject: no printf in drivers code
> 
> Hello,
> 
> We have too many drivers using printf() instead of rte_log().
> Please propose a plan to replace or remove the calls to printf().
> If no plan, I could propose one but you may not like it :)
Does it make sense to add scripts to CI to capture this issue? In general, we do not want printfs in DPDK code (other than tests, sample apps?)

> 
> Affected drivers are:
> 	baseband/acc
> 	bus/dpaa
> 	bus/fslmc
> 	crypto/caam_jr
> 	crypto/ccp
> 	dma/ioat
> 	net/dpaa
> 	net/dpaa2
> 	net/nfp
> 	net/qede
> 	net/txgbe
> 	raw/ifpga
> 
> PS: printf is allowed for tests and dumps.
> 

Re: no printf in drivers code

[Thomas Monjalon]

03/02/2023 17:18, Honnappa Nagarahalli:
> From: Thomas Monjalon <thomas@monjalon.net>
> > 
> > We have too many drivers using printf() instead of rte_log().
> > Please propose a plan to replace or remove the calls to printf().
> > If no plan, I could propose one but you may not like it :)
> 
> Does it make sense to add scripts to CI to capture this issue?
> In general, we do not want printfs in DPDK code (other than tests, sample apps?)

It's difficult to automate because printfs are acceptable
in dump functions, selftest, and testpmd code hosted in drivers directories.

Re: no printf in drivers code

[David Marchand]

On Fri, Feb 3, 2023 at 5:40 PM Thomas Monjalon <thomas@monjalon.net> wrote:
>
> 03/02/2023 17:18, Honnappa Nagarahalli:
> > From: Thomas Monjalon <thomas@monjalon.net>
> > >
> > > We have too many drivers using printf() instead of rte_log().
> > > Please propose a plan to replace or remove the calls to printf().
> > > If no plan, I could propose one but you may not like it :)
> >
> > Does it make sense to add scripts to CI to capture this issue?
> > In general, we do not want printfs in DPDK code (other than tests, sample apps?)
>
> It's difficult to automate because printfs are acceptable
> in dump functions, selftest, and testpmd code hosted in drivers directories.

We can identify which parts need those, create dedicated helpers that
do call printf (but are excluded from the check), and and then the
printf exclusion is applied to the whole tree.


-- 
David Marchand

RE: no printf in drivers code

[Honnappa Nagarahalli]

<snip>

> Subject: Re: no printf in drivers code
> 
> On Fri, Feb 3, 2023 at 5:40 PM Thomas Monjalon <thomas@monjalon.net>
> wrote:
> >
> > 03/02/2023 17:18, Honnappa Nagarahalli:
> > > From: Thomas Monjalon <thomas@monjalon.net>
> > > >
> > > > We have too many drivers using printf() instead of rte_log().
> > > > Please propose a plan to replace or remove the calls to printf().
> > > > If no plan, I could propose one but you may not like it :)
> > >
> > > Does it make sense to add scripts to CI to capture this issue?
> > > In general, we do not want printfs in DPDK code (other than tests,
> > > sample apps?)
> >
> > It's difficult to automate because printfs are acceptable in dump
> > functions, selftest, and testpmd code hosted in drivers directories.
> 
> We can identify which parts need those, create dedicated helpers that do call
> printf (but are excluded from the check), and and then the printf exclusion is
> applied to the whole tree.
+1

> 
> 
> --
> David Marchand

RE: no printf in drivers code

[Walsh, Conor]

> Hello,
> 
> We have too many drivers using printf() instead of rte_log().
> Please propose a plan to replace or remove the calls to printf().
> If no plan, I could propose one but you may not like it :)
> 
> Affected drivers are:
> 	baseband/acc
> 	bus/dpaa
> 	bus/fslmc
> 	crypto/caam_jr
> 	crypto/ccp
> 	dma/ioat
> 	net/dpaa
> 	net/dpaa2
> 	net/nfp
> 	net/qede
> 	net/txgbe
> 	raw/ifpga
> 
> PS: printf is allowed for tests and dumps.

Hi Thomas,

I sent a patch to the mailing list to swap the printf in the IOAT driver to rte log.
I didn't change the fptintf's within the dump function as advised above.

Thanks,
Conor.

RE: no printf in drivers code

[Chautru, Nicolas]

> From: Thomas Monjalon <thomas@monjalon.net>
> 
> Hello,
> 
> We have too many drivers using printf() instead of rte_log().
> Please propose a plan to replace or remove the calls to printf().
> If no plan, I could propose one but you may not like it :)
> 
> Affected drivers are:
> 	baseband/acc

Will push a patch for this next week. These actually currently only occur when called indirectly from test application, but still no concern to replace. 

> 	bus/dpaa
> 	bus/fslmc
> 	crypto/caam_jr
> 	crypto/ccp
> 	dma/ioat
> 	net/dpaa
> 	net/dpaa2
> 	net/nfp
> 	net/qede
> 	net/txgbe
> 	raw/ifpga
> 
> PS: printf is allowed for tests and dumps.
> 

Re: no printf in drivers code

[Niklas Soderlund]

Hi Thomas,

On 2023-02-03 14:57:06 +0100, Thomas Monjalon wrote:
> Hello,
> 
> We have too many drivers using printf() instead of rte_log().
> Please propose a plan to replace or remove the calls to printf().
> If no plan, I could propose one but you may not like it :)
> 
> Affected drivers are:
> 	baseband/acc
> 	bus/dpaa
> 	bus/fslmc
> 	crypto/caam_jr
> 	crypto/ccp
> 	dma/ioat
> 	net/dpaa
> 	net/dpaa2
> 	net/nfp

A patch for NFP have been post as part of the series '[PATCH 0/3] 
cleanup the PMD'.

> 	net/qede
> 	net/txgbe
> 	raw/ifpga
> 
> PS: printf is allowed for tests and dumps.
> 
> 

-- 
Kind Regards,
Niklas Söderlund

RE: no printf in drivers code

[Hemant Agrawal]

 Hi Thomas,
 
 On 2023-02-03 14:57:06 +0100, Thomas Monjalon wrote:
 > Hello,
> >
> > We have too many drivers using printf() instead of rte_log().
> > Please propose a plan to replace or remove the calls to printf().
> > If no plan, I could propose one but you may not like it :)
> >
> > Affected drivers are:
> > 	baseband/acc
> > 	bus/dpaa
> > 	bus/fslmc
> > 	crypto/caam_jr
> > 	crypto/ccp
> > 	dma/ioat
> > 	net/dpaa
> > 	net/dpaa2
[Hemant] NXP will send the patches shortly.


> > 	net/nfp
> > 	net/qede
> > 	net/txgbe
> > 	raw/ifpga
> >
Regards
Hemant