package/rng-tools: make jitterentropy conditional
PATCH 1/1 package/rng-tools: make jitterentropy conditional
The update of rng-tools from 5 to 6.7 introduced a change where the jitterentropy library was enabled by default instead of returning a special 66 return code to hangle the case of no hwrng. This patch reverts that change and allows a user to select when to enable the jitterentropy source. The bug documents an issue of when a hwrng is enabled with jitterentropy there is a longer boot time.
Fixes:
https://bugs.busybox.net/show_bug.cgi?id=12511
--- package/rng-tools/Config.in | 1 - package/rng-tools/rng-tools.mk | 9 ++++++++- package/rng-tools/rngd.service | 1 + 3 files changed, 9 insertions(+), 2 deletions(-)
package/rng-tools/Config.in +0 -1
@@ -4,7 +4,6 @@ config BR2_PACKAGE_RNG_TOOLS # pthread_setaffinity_np depends on BR2_TOOLCHAIN_HAS_THREADS_NPTL select BR2_PACKAGE_ARGP_STANDALONE if BR2_TOOLCHAIN_USES_UCLIBC || BR2_TOOLCHAIN_USES_MUSL - select BR2_PACKAGE_JITTERENTROPY_LIBRARY # For rdrand & darn ligcrypt is required and it's not obvious to users select BR2_PACKAGE_LIBGCRYPT if BR2_i386 || BR2_x86_64 || BR2_powerpc64le select BR2_PACKAGE_LIBSYSFS
package/rng-tools/rng-tools.mk +8 -1
@@ -8,7 +8,7 @@ RNG_TOOLS_VERSION = 6.8 RNG_TOOLS_SITE = $(call github,nhorman,$(RNG_TOOLS_NAME),v$(RNG_TOOLS_VERSION)) RNG_TOOLS_LICENSE = GPL-2.0 RNG_TOOLS_LICENSE_FILES = COPYING -RNG_TOOLS_DEPENDENCIES = libsysfs jitterentropy-library host-pkgconf +RNG_TOOLS_DEPENDENCIES = libsysfs host-pkgconf # From git RNG_TOOLS_AUTORECONF = YES
@@ -29,6 +29,13 @@ else RNG_TOOLS_CONF_OPTS += --without-libgcrypt endif +ifeq ($(BR2_PACKAGE_JITTERENTROPY_LIBRARY),y) +RNG_TOOLS_DEPENDENCIES += jitterentropy-library +RNG_TOOLS_CONF_OPTS += --enable-jitterentropy +else +RNG_TOOLS_CONF_OPTS += --disable-jitterentropy +endif + define RNG_TOOLS_INSTALL_INIT_SYSV $(INSTALL) -D -m 755 package/rng-tools/S21rngd \ $(TARGET_DIR)/etc/init.d/S21rngd
package/rng-tools/rngd.service +1 -0
@@ -3,6 +3,7 @@ Description=Hardware RNG Entropy Gatherer Daemon [Service] ExecStart=/usr/sbin/rngd -f $DAEMON_ARGS +SuccessExitStatus=66 EnvironmentFile=-/etc/default/rngd [Install]
Signature · 1 line
2.17.1
PATCH 1/1 package/rng-tools: make jitterentropy conditional
Quoted 7 lines · On Wed, 5 Feb 2020 12:26:08 -0600 Matt Weber…
The update of rng-tools from 5 to 6.7 introduced a change where the jitterentropy library was enabled by default instead of returning a special 66 return code to hangle the case of no hwrng. This patch reverts that change and allows a user to select when to enable the jitterentropy source. The bug documents an issue of when a hwrng is enabled with jitterentropy there is a longer boot time.
I don't understand this story of the 66 return code. Could you explain a bit more ?
package/rng-tools/rngd.service Description=Hardware RNG Entropy Gatherer Daemon · line 6
Quoted 8 lines
index
11386d1e5d..c0bcffe59e100644package/rng-tools/rngd.service +1
-0@@ -3,6 +3,7 @@ Description=Hardware RNG Entropy Gatherer Daemon [Service] ExecStart=/usr/sbin/rngd -f $DAEMON_ARGS +SuccessExitStatus=66
Will this work even if jitterentropy support is enabled ?
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Bootlin Embedded Linux and Kernel engineering https://bootlin.com
PATCH 1/1 package/rng-tools: make jitterentropy conditional
Thomas,
Quoted 12 lines · Thomas Petazzoni
Quoted 7 lines · On Wed, 5 Feb 2020 12:26:08 -0600 Matt Weber…
The update of rng-tools from 5 to 6.7 introduced a change where the jitterentropy library was enabled by default instead of returning a special 66 return code to hangle the case of no hwrng. This patch reverts that change and allows a user to select when to enable the jitterentropy source. The bug documents an issue of when a hwrng is enabled with jitterentropy there is a longer boot time.
Quoted 2 lines
I don't understand this story of the 66 return code. Could you explain a bit more ?
In Buildroot commit 22cb51e1 the systemd support addition included the successcode change based on the Fedora bugfix for ignoring if no hwrng is present (https://bugzilla.redhat.com/show_bug.cgi?id=892178).
Quoted 11 lines
package/rng-tools/rngd.service Description=Hardware RNG Entropy Gatherer Daemon · line 6
Quoted 9 lines
index
11386d1e5d..c0bcffe59e100644package/rng-tools/rngd.service +1
-0@@ -3,6 +3,7 @@ Description=Hardware RNG Entropy Gatherer Daemon [Service] ExecStart=/usr/sbin/rngd -f $DAEMON_ARGS +SuccessExitStatus=66Will this work even if jitterentropy support is enabled ?
My understanding is that you'd never hit this return case when jitterentropy support built in as it would always init the entropy pool using that library as a source. Thus the problem of the boot delay even when hwrng is present.
Matt
PATCH 1/1 package/rng-tools: make jitterentropy conditional
Thomas,
Quoted 17 lines · Matthew Weber
Thomas,
Quoted 12 lines · Thomas Petazzoni
Quoted 7 lines · On Wed, 5 Feb 2020 12:26:08 -0600 Matt Weber…
The update of rng-tools from 5 to 6.7 introduced a change where the jitterentropy library was enabled by default instead of returning a special 66 return code to hangle the case of no hwrng. This patch reverts that change and allows a user to select when to enable the jitterentropy source. The bug documents an issue of when a hwrng is enabled with jitterentropy there is a longer boot time.
Quoted 6 lines
Quoted 2 lines
I don't understand this story of the 66 return code. Could you explain a bit more ?
In Buildroot commit
22cb51e1the systemd support addition included the successcode change based on the Fedora bugfix for ignoring if no hwrng is present (https://bugzilla.redhat.com/show_bug.cgi?id=892178).
Ryan Barnett noticed that rng-tools dropped the special return code. I guess we could just drop support for this special case as well....
https://github.com/nhorman/rng-tools/blob/v6.8/rngd.c#L805
Quoted 19 lines
Quoted 7 lines
Quoted 7 lines
index
11386d1e5d..c0bcffe59e100644package/rng-tools/rngd.service +0
-0@@ -3,6 +3,7 @@ Description=Hardware RNG Entropy Gatherer DaemonQuoted 5 lines
package/rng-tools/rngd.service Description=Hardware RNG Entropy Gatherer Daemon · line 6
Quoted 3 lines
package/rng-tools/rngd.service +1
-0[Service] ExecStart=/usr/sbin/rngd -f $DAEMON_ARGS +SuccessExitStatus=66Will this work even if jitterentropy support is enabled ?
My understanding is that you'd never hit this return case when jitterentropy support built in as it would always init the entropy pool using that library as a source. Thus the problem of the boot delay even when hwrng is present.
Matt
Signature · 23 lines
Matthew Weber | Associate Director Software Engineer | Commercial Avionics COLLINS AEROSPACE 400 Collins Road NE, Cedar Rapids, Iowa 52498, USA Tel: +1 319 295 7349 | FAX: +1 319 263 6099 matthew.weber at collins.com | collinsaerospace.com CONFIDENTIALITY WARNING: This message may contain proprietary and/or privileged information of Collins Aerospace and its affiliated companies. If you are not the intended recipient, please 1) Do not disclose, copy, distribute or use this message or its contents. 2) Advise the sender by return email. 3) Delete all copies (including all attachments) from your computer. Your cooperation is greatly appreciated. Any export restricted material should be shared using my matthew.weber at corp.rockwellcollins.com address.
PATCH 1/1 package/rng-tools: make jitterentropy conditional
Quoted 25 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
"Matthew" == Matthew Weber <[email protected]> writes:
Thomas,
Quoted 20 lines · Matthew Weber
Thomas,
Quoted 14 lines · Thomas Petazzoni
Quoted 7 lines · On Wed, 5 Feb 2020 12:26:08 -0600 Matt Weber…
The update of rng-tools from 5 to 6.7 introduced a change where the jitterentropy library was enabled by default instead of returning a special 66 return code to hangle the case of no hwrng. This patch reverts that change and allows a user to select when to enable the jitterentropy source. The bug documents an issue of when a hwrng is enabled with jitterentropy there is a longer boot time.
I don't understand this story of the 66 return code. Could you explain a bit more ?
Quoted 7 lines
Quoted 3 lines
In Buildroot commit
22cb51e1the systemd support addition included the successcode change based on the Fedora bugfix for ignoring if no hwrng is present (https://bugzilla.redhat.com/show_bug.cgi?id=892178).Ryan Barnett noticed that rng-tools dropped the special return code. I guess we could just drop support for this special case as well....
https://github.com/nhorman/rng-tools/blob/v6.8/rngd.c#L805
Has that Fedora patch ever been included upstream? I don't seem to find it in the history.
I btw see there is a 6.9 release with more bugfixes, care to send a patch bumping the version?
Signature · 1 line
Bye, Peter Korsgaard
PATCH 1/1 package/rng-tools: make jitterentropy conditional
Peter,
Quoted 30 lines · Peter Korsgaard
Quoted 29 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
"Matthew" == Matthew Weber <[email protected]> writes:
Thomas,
Quoted 23 lines · Matthew Weber
Thomas,
Quoted 12 lines · Thomas Petazzoni
Quoted 7 lines · On Wed, 5 Feb 2020 12:26:08 -0600 Matt Weber…
The update of rng-tools from 5 to 6.7 introduced a change where the jitterentropy library was enabled by default instead of returning a special 66 return code to hangle the case of no hwrng. This patch reverts that change and allows a user to select when to enable the jitterentropy source. The bug documents an issue of when a hwrng is enabled with jitterentropy there is a longer boot time.
Quoted 2 lines
I don't understand this story of the 66 return code. Could you explain a bit more ?
In Buildroot commit
22cb51e1the systemd support addition included the successcode change based on the Fedora bugfix for ignoring if no hwrng is present (https://bugzilla.redhat.com/show_bug.cgi?id=892178).
Quoted 6 lines
Quoted 3 lines
Ryan Barnett noticed that rng-tools dropped the special return code. I guess we could just drop support for this special case as well....
https://github.com/nhorman/rng-tools/blob/v6.8/rngd.c#L805Has that Fedora patch ever been included upstream? I don't seem to find it in the history.
It doesn't look like it. I'm wondering if we should just drop this systemd unit return behavior and put a comment in the service file about enabling an entropy package if this service fails without a hardware rng present. Seems too complicated to try and cover this special case.
Quoted 3 lines
I btw see there is a 6.9 release with more bugfixes, care to send a patch bumping the version?
I'll review the one that was just sent. I also ping'd the original bug report to see if they could test with this bump as there are a number of minor jitterentropy related fixes.
http://patchwork.ozlabs.org/patch/1235396/
Matt
PATCH 1/1 package/rng-tools: make jitterentropy conditional
Hi Matt,
Quoted 49 lines · Matthew Weber
Peter,
Quoted 30 lines · Peter Korsgaard
Quoted 29 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
"Matthew" == Matthew Weber <[email protected]> writes:
Thomas,
Quoted 23 lines · Matthew Weber
Thomas,
Quoted 12 lines · Thomas Petazzoni
Quoted 7 lines · On Wed, 5 Feb 2020 12:26:08 -0600 Matt Weber…
The update of rng-tools from 5 to 6.7 introduced a change where the jitterentropy library was enabled by default instead of returning a special 66 return code to hangle the case of no hwrng. This patch reverts that change and allows a user to select when to enable the jitterentropy source. The bug documents an issue of when a hwrng is enabled with jitterentropy there is a longer boot time.
Quoted 2 lines
I don't understand this story of the 66 return code. Could you explain a bit more ?
In Buildroot commit
22cb51e1the systemd support addition included the successcode change based on the Fedora bugfix for ignoring if no hwrng is present (https://bugzilla.redhat.com/show_bug.cgi?id=892178).Quoted 6 lines
Quoted 3 lines
Ryan Barnett noticed that rng-tools dropped the special return code. I guess we could just drop support for this special case as well....
https://github.com/nhorman/rng-tools/blob/v6.8/rngd.c#L805Has that Fedora patch ever been included upstream? I don't seem to find it in the history.
It doesn't look like it. I'm wondering if we should just drop this systemd unit return behavior and put a comment in the service file about enabling an entropy package if this service fails without a hardware rng present. Seems too complicated to try and cover this special case.
Quoted 1 line
Quoted 7 lines
Quoted 2 lines
I btw see there is a 6.9 release with more bugfixes, care to send a patch bumping the version?
I'll review the one that was just sent. I also ping'd the original bug report to see if they could test with this bump as there are a number of minor jitterentropy related fixes.
http://patchwork.ozlabs.org/patch/1235396/
We had similar issue today with am335x (kernel 5.4.x). Bumping rng-tools to v6.9 helped.
The next bump will be more interesting as rng-tools moves to openssl instead of libgcrypt and this dependency is mandatory.
Yegor
PATCH 1/1 package/rng-tools: make jitterentropy conditional
Yegor,
Quoted 57 lines · Yegor Yefremov
Hi Matt,
Quoted 52 lines · Matthew Weber
Peter,
Quoted 30 lines · Peter Korsgaard
Quoted 29 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
"Matthew" == Matthew Weber <[email protected]> writes:
Thomas,
Quoted 23 lines · Matthew Weber
Thomas,
Quoted 12 lines · Thomas Petazzoni
Quoted 7 lines · On Wed, 5 Feb 2020 12:26:08 -0600 Matt Weber…
The update of rng-tools from 5 to 6.7 introduced a change where the jitterentropy library was enabled by default instead of returning a special 66 return code to hangle the case of no hwrng. This patch reverts that change and allows a user to select when to enable the jitterentropy source. The bug documents an issue of when a hwrng is enabled with jitterentropy there is a longer boot time.
Quoted 2 lines
I don't understand this story of the 66 return code. Could you explain a bit more ?
In Buildroot commit
22cb51e1the systemd support addition included the successcode change based on the Fedora bugfix for ignoring if no hwrng is present (https://bugzilla.redhat.com/show_bug.cgi?id=892178).Quoted 6 lines
Quoted 3 lines
Ryan Barnett noticed that rng-tools dropped the special return code. I guess we could just drop support for this special case as well....
https://github.com/nhorman/rng-tools/blob/v6.8/rngd.c#L805Has that Fedora patch ever been included upstream? I don't seem to find it in the history.
It doesn't look like it. I'm wondering if we should just drop this systemd unit return behavior and put a comment in the service file about enabling an entropy package if this service fails without a hardware rng present. Seems too complicated to try and cover this special case.
Quoted 3 lines
I btw see there is a 6.9 release with more bugfixes, care to send a patch bumping the version?
Quoted 8 lines
Quoted 4 lines
I'll review the one that was just sent. I also ping'd the original bug report to see if they could test with this bump as there are a number of minor jitterentropy related fixes.
http://patchwork.ozlabs.org/patch/1235396/We had similar issue today with am335x (kernel 5.4.x). Bumping rng-tools to v6.9 helped.
Thank you for that feedback, I'll make this patch is superseded as the bump resolved the bug. I'll add a note with this patchwork thread in the bug report as well.
Matt
PATCH 1/1 package/rng-tools: make jitterentropy conditional
Hello Matt,
Quoted 6 lines · On Thu, 13 Feb 2020 11:07:36 -0600 Matthew Weber…
Quoted 2 lines
We had similar issue today with am335x (kernel 5.4.x). Bumping rng-tools to v6.9 helped.
Thank you for that feedback, I'll make this patch is superseded as the bump resolved the bug. I'll add a note with this patchwork thread in the bug report as well.
As I was going through open bug reports, I looked again at
https://bugs.busybox.net/show_bug.cgi?id=12511, and I disagree with the conclusion of discarding this patch "package/rng-tools: make jitterentropy conditional".
If I understand correctly:
- If you have a hardware RNG, the jitterentropy library is not necessary.
- If you don't have a hardware RNG, rng-tools will error out, unless jitterentropy support is enabled.
So, we really want to make the jitterentropy support in rng-tools optional, and leave it up to the user to enable it if (s)he has no hardware RNG available. And if there's no hardware RNG and no jitterentropy support, rng-tools will error out, the service will fail to start, and that's good. So the trick with the special 66 return code is not necessary.
Do you agree ?
Best regards,
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Bootlin Embedded Linux and Kernel engineering https://bootlin.com
PATCH 1/1 package/rng-tools: make jitterentropy conditional
Thomas,
Quoted 18 lines · Thomas Petazzoni
Hello Matt,
Quoted 6 lines · On Thu, 13 Feb 2020 11:07:36 -0600 Matthew Weber…
Quoted 2 lines
We had similar issue today with am335x (kernel 5.4.x). Bumping rng-tools to v6.9 helped.
Thank you for that feedback, I'll make this patch is superseded as the bump resolved the bug. I'll add a note with this patchwork thread in the bug report as well.
As I was going through open bug reports, I looked again at
https://bugs.busybox.net/show_bug.cgi?id=12511, and I disagree with the conclusion of discarding this patch "package/rng-tools: make jitterentropy conditional".
Quoted 5 lines
If I understand correctly:
- If you have a hardware RNG, the jitterentropy library is not necessary.
Right, jitterentropy is a fall back if there isn't a hardware RNG setup. The performance impacts of having it enabled when there is a hardware RNG were improved between versions and that was the motivation to drop this patch as it didn't seem to still be a bug (having it enabled when it isn't really needed).
Quoted 2 lines
- If you don't have a hardware RNG, rng-tools will error out, unless jitterentropy support is enabled.
Correct.
Quoted 9 lines
So, we really want to make the jitterentropy support in rng-tools optional, and leave it up to the user to enable it if (s)he has no hardware RNG available. And if there's no hardware RNG and no jitterentropy support, rng-tools will error out, the service will fail to start, and that's good. So the trick with the special 66 return code is not necessary.
Do you agree ?
I agree, we should drop that rngd.service change but still keep the option where a user could add the jitterentropy library if they need it.
Regards,
Matt