sysfs output without newlines

sysfs output without newlines

[Joe Perches]

While doing an investigation for a possible treewide conversion of
sysfs output using sprintf/snprintf/scnprintf, I discovered
several instances of sysfs output without terminating newlines.

It seems likely all of these should have newline terminations
or have the \n\r termination changed to a single newline.

Anyone have any objection to patches adding newlines to these
in their original forms using sprintf/snprintf/scnprintf?

A few of these might be false positives as
	"%s", string
might already have string with a newline termination.

+++ drivers/edac/fsl_ddr_edac.c
+	return sysfs_emit(data, "0x%08x",
+	return sysfs_emit(data, "0x%08x",
+	return sysfs_emit(data, "0x%08x",
+++ drivers/edac/synopsys_edac.c
+	return sysfs_emit(data, "Data Poisoning: %s\n\r",
+		return sysfs_emit(buf, "1");
+		return sysfs_emit(buf, "0");
+		return sysfs_emit(buf, "1");
+		return sysfs_emit(buf, "0");
+		return sysfs_emit(buf, "0");
+		return sysfs_emit(buf, "1");
+		return sysfs_emit(buf, "0");
+	return sysfs_emit(buf, "%u", !!(data->status & mask));
+	return sysfs_emit(buf, "%u", data->tcrit2[index] * 1000);
+	return sysfs_emit(buf, "%d",
+	return sysfs_emit(buf, "%u", data->tcrit1[index] * 1000);
+	return sysfs_emit(buf, "%d",
+	return sysfs_emit(buf, "%d", data->toffset[index] * 500);
+++ drivers/i3c/master.c
+		return sysfs_emit(buf, "i3c:dcr%02Xmanuf%04X", devinfo.dcr,
+	return sysfs_emit(buf, "i3c:dcr%02Xmanuf%04Xpart%04Xext%04X",
+	return sysfs_emit(buf, "%s", dd->boardversion);
+	return sysfs_emit(buf, "%s", dd->serial);
+	return sysfs_emit(buf, "%s", (char *)ib_qib_version);
+	return sysfs_emit(buf, "%s", dd->boardversion);
+	return sysfs_emit(buf, "%s", dd->lbus_info);
+	return sysfs_emit(buf, "ipac:f%02Xv%08Xd%08X", idev->id_format,
+++ drivers/memstick/core/mspro_block.c
+	return sysfs_emit(buffer, "%s", (char *)s_attr->data);
+	return sysfs_emit(buf, "%s",
+	return sysfs_emit(buf, "%s",
+++ drivers/misc/mei/bus.c
+	return sysfs_emit(buf, "%s", cldev->name);
+	return sysfs_emit(buf, "%pUl", uuid);
+	return sysfs_emit(buf, "%02X", version);
+	return sysfs_emit(buf, "mei:%s:%pUl:%02X:",
+	return sysfs_emit(buf, "%d", maxconn);
+	return sysfs_emit(buf, "%d", fixed);
+	return sysfs_emit(buf, "%d", vt);
+	return sysfs_emit(buf, "%u", maxlen);
+	return sysfs_emit(buf, "%s", mei_dev_state_str(dev_state));
+++ drivers/misc/tifm_core.c
+	return sysfs_emit(buf, "%x", sock->type);
+			return sysfs_emit(buf, "%s",
+	return sysfs_emit(buf, "%d", dev->net_count);
+++ drivers/net/ethernet/freescale/dpaa/dpaa_eth_sysfs.c
+		return sysfs_emit(buf, "%llx",
+		return sysfs_emit(buf, "none");
+++ drivers/net/ethernet/ibm/ehea/ehea_main.c
+	return sysfs_emit(buf, "%d", port->logical_port_id);
+++ drivers/net/wireless/intel/ipw2x00/ipw2200.c
+		return sysfs_emit(buf, "%s", priv->prom_net_dev->name);
+	return sysfs_emit(buf, "0x%04X",
+	return sysfs_emit(buf, "%d", il->retry_rate);
+	return sysfs_emit(buf, "%pOF", np);
+	return sysfs_emit(buf, "pcmcia:m%04Xc%04Xf%02Xfn%02Xpfn%02X"
+++ drivers/platform/x86/dell-smbios-base.c
+		return sysfs_emit(buf, "%08x", da_tokens[i].location);
+		return sysfs_emit(buf, "%08x", da_tokens[i].value);
+	return sysfs_emit(buf, "%08x",
+++ drivers/scsi/st.c
+	return sysfs_emit(buf, "%lld",
+	return sysfs_emit(buf, "%lld",
+	return sysfs_emit(buf, "%lld",
+	return sysfs_emit(buf, "%lld",
+	return sysfs_emit(buf, "%lld",
+	return sysfs_emit(buf, "%lld",
+	return sysfs_emit(buf, "%lld",
+	return sysfs_emit(buf, "%lld",
+	return sysfs_emit(buf, "%lld",
+	return sysfs_emit(buf, "%lld",
+++ drivers/spi/spi-tle62x0.c
+	return sysfs_emit(buf, "%d", value);
+++ drivers/usb/class/cdc-acm.c
+	return sysfs_emit(buf, "%d", acm->ctrl_caps);
+	return sysfs_emit(buf, "%d", acm->country_rel_date);
+++ drivers/usb/class/usblp.c
+	return sysfs_emit(buf, "%s", usblp->device_id_string+2);
+	return sysfs_emit(buf, "usb:v%04Xp%04Xd%04Xdc%02Xdsc%02Xdp%02X"
+++ drivers/usb/misc/cytherm.c
+	return sysfs_emit(buf, "%i", cytherm->brightness);
+	return sysfs_emit(buf, "%c%i.%i", sign ? '-' : '+', temp >> 1,
+		return sysfs_emit(buf, "1");
+		return sysfs_emit(buf, "0");
+	return sysfs_emit(buf, "%d", retval);
+	return sysfs_emit(buf, "%d", retval);

Re: sysfs output without newlines

[Denis Efremov]

Hi,

On 8/29/20 9:23 PM, Joe Perches wrote:
> While doing an investigation for a possible treewide conversion of
> sysfs output using sprintf/snprintf/scnprintf, I discovered
> several instances of sysfs output without terminating newlines.
> 
> It seems likely all of these should have newline terminations
> or have the \n\r termination changed to a single newline.

I think that it could break badly written scripts in rare cases.

> 
> Anyone have any objection to patches adding newlines to these
> in their original forms using sprintf/snprintf/scnprintf?

I'm not sure about existing cases, but I think it's a good
checkpatch.pl warning for new patches. It should be 
possible to check sysfs_emit() calls.

Thanks,
Denis

Re: sysfs output without newlines

[Joe Perches]

On Sat, 2020-08-29 at 23:23 +0300, Denis Efremov wrote:
> Hi,
> 
> On 8/29/20 9:23 PM, Joe Perches wrote:
> > While doing an investigation for a possible treewide conversion of
> > sysfs output using sprintf/snprintf/scnprintf, I discovered
> > several instances of sysfs output without terminating newlines.
> > 
> > It seems likely all of these should have newline terminations
> > or have the \n\r termination changed to a single newline.
> 
> I think that it could break badly written scripts in rare cases.

Maybe.

Is sysfs output a nominally unchangeable api like seq_?
Dunno.  seq_ output is extended all the time.

I think whitespace isn't generally considered part of
sscanf type input content awareness.

> > Anyone have any objection to patches adding newlines to these
> > in their original forms using sprintf/snprintf/scnprintf?
> 
> I'm not sure about existing cases, but I think it's a good
> checkpatch.pl warning for new patches. It should be 
> possible to check sysfs_emit() calls.

Eventually, yes.

cheers, Joe

Re: sysfs output without newlines

[Greg Kroah-Hartman]

On Sat, Aug 29, 2020 at 11:23:43AM -0700, Joe Perches wrote:
> While doing an investigation for a possible treewide conversion of
> sysfs output using sprintf/snprintf/scnprintf, I discovered
> several instances of sysfs output without terminating newlines.
> 
> It seems likely all of these should have newline terminations
> or have the \n\r termination changed to a single newline.
> 
> Anyone have any objection to patches adding newlines to these
> in their original forms using sprintf/snprintf/scnprintf?

No objection for me, patches for my subsystems will be gladly taken.

thanks,

greg k-h

RE: sysfs output without newlines

[David Laight]

From: Joe Perches
> Sent: 29 August 2020 21:34
...
> > On 8/29/20 9:23 PM, Joe Perches wrote:
> > > While doing an investigation for a possible treewide conversion of
> > > sysfs output using sprintf/snprintf/scnprintf, I discovered
> > > several instances of sysfs output without terminating newlines.
> > >
> > > It seems likely all of these should have newline terminations
> > > or have the \n\r termination changed to a single newline.
> >
> > I think that it could break badly written scripts in rare cases.
> 
> Maybe.
> 
> Is sysfs output a nominally unchangeable api like seq_?
> Dunno.  seq_ output is extended all the time.
> 
> I think whitespace isn't generally considered part of
> sscanf type input content awareness.

The shell will remove trailing '\n' (but not '\r') from:
	foo=$(cat bar)
So shell scripts are unlikely to be affected.

	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)

Re: sysfs output without newlines

[Pete Zaitcev]

On Tue, 1 Sep 2020 09:07:42 +0000
David Laight <David.Laight@ACULAB.COM> wrote:

> The shell will remove trailing '\n' (but not '\r') from:
> 	foo=$(cat bar)
> So shell scripts are unlikely to be affected.

Surprisingly enough, Python is okay with this:

    with open("/proc/sys/net/ipv4/ip_forward",'r') as fp:
        print(int(fp.read()))

Nonetheless, I would rather no add those newlines. I saw enough
weird software that is sensitive where it's should not bee.

-- Pete

Re: sysfs output without newlines

[Andy Shevchenko]

On Wed, Sep 2, 2020 at 1:24 AM Pete Zaitcev <zaitcev@redhat.com> wrote:
>
> On Tue, 1 Sep 2020 09:07:42 +0000
> David Laight <David.Laight@ACULAB.COM> wrote:
>
> > The shell will remove trailing '\n' (but not '\r') from:
> >       foo=$(cat bar)
> > So shell scripts are unlikely to be affected.
>
> Surprisingly enough, Python is okay with this:
>
>     with open("/proc/sys/net/ipv4/ip_forward",'r') as fp:
>         print(int(fp.read()))
>
> Nonetheless, I would rather no add those newlines. I saw enough
> weird software that is sensitive where it's should not bee.

There are still a lot of scripts in PHP and Perl. Besides that I think
the many users of such files are in clouds, means Go should be checked
even before these.

-- 
With Best Regards,
Andy Shevchenko