[PATCH] powerpc/powernv: Make possible for user to force a full ipl cec reboot

Ever since fast reboot is enabled by default in opal,
opal_cec_reboot() will use fast-reset instead of full IPL to perform
system reboot. This leaves the user with no direct way to force a full
IPL reboot except changing an nvram setting that persistently disables
fast-reset for all subsequent reboots.

This patch provides a more direct way for the user to force a one-shot
full IPL reboot by passing the command line argument 'full' to the
reboot command. So the user will be able to tweak the reboot behavior
via:

  $ sudo reboot full	# Force a full ipl reboot skipping fast-reset

  or
  $ sudo reboot  	# default reboot path (usually fast-reset)

The reboot command passes the un-parsed command argument to the kernel
via the 'Reboot' syscall which is then passed on to the arch function
pnv_restart(). The patch updates pnv_restart() to handle this cmd-arg
and issues opal_cec_reboot2 with OPAL_REBOOT_FULL_IPL to force a full
IPL reset.

Signed-off-by: Vaibhav Jain <redacted>
---
 arch/powerpc/include/asm/opal-api.h    | 1 +
 arch/powerpc/platforms/powernv/setup.c | 8 +++++++-
 2 files changed, 8 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/include/asm/opal-api.h b/arch/powerpc/include/asm/opal-api.h
index 8365353330b4..870fb7b239ea 100644
--- a/arch/powerpc/include/asm/opal-api.h
+++ b/arch/powerpc/include/asm/opal-api.h

@@ -1050,6 +1050,7 @@ enum OpalSysCooling {
 enum {
 	OPAL_REBOOT_NORMAL		= 0,
 	OPAL_REBOOT_PLATFORM_ERROR	= 1,
+	OPAL_REBOOT_FULL_IPL		= 2,
 };
 
 /* Argument to OPAL_PCI_TCE_KILL */

diff --git a/arch/powerpc/platforms/powernv/setup.c b/arch/powerpc/platforms/powernv/setup.c
index adddde023622..33d2faeacff8 100644
--- a/arch/powerpc/platforms/powernv/setup.c
+++ b/arch/powerpc/platforms/powernv/setup.c

@@ -224,7 +224,13 @@ static void  __noreturn pnv_restart(char *cmd)
 	pnv_prepare_going_down();
 
 	while (rc == OPAL_BUSY || rc == OPAL_BUSY_EVENT) {
-		rc = opal_cec_reboot();
+
+		/* See if we need to do a full IPL reboot */
+		if (cmd && strcmp(cmd, "full") == 0)
+			rc = opal_cec_reboot2(OPAL_REBOOT_FULL_IPL, NULL);
+		else
+			rc = opal_cec_reboot();
+
 		if (rc == OPAL_BUSY_EVENT)
 			opal_poll_events(NULL);
 		else

-- 
2.17.1

On Sat,  1 Sep 2018 13:47:45 +0530
Vaibhav Jain [off-list ref] wrote:

Ever since fast reboot is enabled by default in opal,
opal_cec_reboot() will use fast-reset instead of full IPL to perform
system reboot. This leaves the user with no direct way to force a full
IPL reboot except changing an nvram setting that persistently disables
fast-reset for all subsequent reboots.

This patch provides a more direct way for the user to force a one-shot
full IPL reboot by passing the command line argument 'full' to the
reboot command. So the user will be able to tweak the reboot behavior
via:

  $ sudo reboot full	# Force a full ipl reboot skipping fast-reset

  or
  $ sudo reboot  	# default reboot path (usually fast-reset)

The reboot command passes the un-parsed command argument to the kernel
via the 'Reboot' syscall which is then passed on to the arch function
pnv_restart(). The patch updates pnv_restart() to handle this cmd-arg
and issues opal_cec_reboot2 with OPAL_REBOOT_FULL_IPL to force a full
IPL reset.

Signed-off-by: Vaibhav Jain <redacted>

This is nice but I wonder if the default should be IPL and fast
should be the option.

Thanks for looking into this patch Nick,

Nicholas Piggin [off-list ref] writes:

This is nice but I wonder if the default should be IPL and fast
should be the option.

Fast-reset should work most of the times so I think it should be
default. Its also the current default which I am bit relunctant to
change.

We usually need full IPL reset only when we need to update skiboot or
have malfunctioning device that isnt being reset properly during
fast-reset. So I am believe full IPL resets will be less frequent hence
should be explicitly requested.

-- 
Vaibhav Jain [off-list ref]
Linux Technology Center, IBM India Pvt. Ltd.

On 01/09/18 18:17, Vaibhav Jain wrote:

Ever since fast reboot is enabled by default in opal,
opal_cec_reboot() will use fast-reset instead of full IPL to perform
system reboot. This leaves the user with no direct way to force a full
IPL reboot except changing an nvram setting that persistently disables
fast-reset for all subsequent reboots.
=20
This patch provides a more direct way for the user to force a one-shot
full IPL reboot by passing the command line argument 'full' to the
reboot command. So the user will be able to tweak the reboot behavior
via:
=20
   $ sudo reboot full	# Force a full ipl reboot skipping fast-reset
=20
   or
   $ sudo reboot  	# default reboot path (usually fast-reset)
=20
The reboot command passes the un-parsed command argument to the kernel
via the 'Reboot' syscall which is then passed on to the arch function
pnv_restart(). The patch updates pnv_restart() to handle this cmd-arg
and issues opal_cec_reboot2 with OPAL_REBOOT_FULL_IPL to force a full
IPL reset.
=20
Signed-off-by: Vaibhav Jain <redacted>

Oh good, someone else has finally picked this up and sent a kernel=20
patch, I did the skiboot half and then neglected to make it useful (I=20
sent an RFC at https://patchwork.ozlabs.org/patch/697604/ but never=20
followed up on it... this approach seems more usable, I think).

When you say "the reboot command" - is this behaviour of passing the=20
argument common to all the important init systems/reboot utils?  What's=20
the correct systemd way to do it?

  	while (rc =3D=3D OPAL_BUSY || rc =3D=3D OPAL_BUSY_EVENT) {
-		rc =3D opal_cec_reboot();
+
+		/* See if we need to do a full IPL reboot */
+		if (cmd && strcmp(cmd, "full") =3D=3D 0)
+			rc =3D opal_cec_reboot2(OPAL_REBOOT_FULL_IPL, NULL);

You might want to check for OPAL_UNSUPPORTED here just in case we're=20
running on ancient firmware.

+		else
+			rc =3D opal_cec_reboot();
+
  		if (rc =3D=3D OPAL_BUSY_EVENT)
  			opal_poll_events(NULL);
  		else
=20

--=20
Andrew Donnellan              OzLabs, ADL Canberra
andrew.donnellan@au1.ibm.com  IBM Australia Limited

Thanks for looking into this Andrew,

Andrew Donnellan [off-list ref] writes:

Oh good, someone else has finally picked this up and sent a kernel 
patch, I did the skiboot half and then neglected to make it useful (I 
sent an RFC at https://patchwork.ozlabs.org/patch/697604/ but never 
followed up on it... this approach seems more usable, I think).

Thanks for pointing that out. I wasnt aware of this patch.

When you say "the reboot command" - is this behaviour of passing the 
argument common to all the important init systems/reboot utils?  What's 
the correct systemd way to do it?

Mostly Yes,

On systemd the reboot command is just a symlink to systemctl binary that
checks argv[0] and maintains the a backward compatible behaviour. The
more systemd specific command 'systemctl reboot full' will also do the
same thing. So this patch should work on systemd.

Looking at 'upstart' I see support for adding an arg to the reboot
command since revision
https://bazaar.launchpad.net/~upstart-devel/upstart/trunk/revision/1432.3.1
. So upstart should also be supported.

I dont see support for sending an arg in 'SysVinit' halt command
yet. However this patch wont break the halt command. Just that SysVinit
systems will continue to use fast-reboot like today. Though
theorectially with this patch a SysVinit user can write a tool to issue
reboot syscall with cmd == LINUX_REBOOT_CMD_RESTART2 and arg == "full".

quoted

  	while (rc == OPAL_BUSY || rc == OPAL_BUSY_EVENT) {
-		rc = opal_cec_reboot();
+
+		/* See if we need to do a full IPL reboot */
+		if (cmd && strcmp(cmd, "full") == 0)
+			rc = opal_cec_reboot2(OPAL_REBOOT_FULL_IPL, NULL);

You might want to check for OPAL_UNSUPPORTED here just in case we're 
running on ancient firmware.

Good idea, will fix this in v2.

-- 
Vaibhav Jain [off-list ref]
Linux Technology Center, IBM India Pvt. Ltd.

Vaibhav Jain [off-list ref] writes:

Ever since fast reboot is enabled by default in opal,
opal_cec_reboot() will use fast-reset instead of full IPL to perform
system reboot. This leaves the user with no direct way to force a full
IPL reboot except changing an nvram setting that persistently disables
fast-reset for all subsequent reboots.

This patch provides a more direct way for the user to force a one-shot
full IPL reboot by passing the command line argument 'full' to the
reboot command. So the user will be able to tweak the reboot behavior
via:

  $ sudo reboot full	# Force a full ipl reboot skipping fast-reset

  or
  $ sudo reboot  	# default reboot path (usually fast-reset)

The reboot command passes the un-parsed command argument to the kernel
via the 'Reboot' syscall which is then passed on to the arch function
pnv_restart(). The patch updates pnv_restart() to handle this cmd-arg
and issues opal_cec_reboot2 with OPAL_REBOOT_FULL_IPL to force a full
IPL reset.

We're about to introduce an MPIPL reboot type (to take a firmware
assisted kdump style thing), and we maybe should have a reboot type to
force attempting a fast-reboot, and this makes me think if we should add
those in now?

If the reboot type isn't supported, what should be the behvaiour? Reboot
the default way or don't reboot at all?

-- 
Stewart Smith
OPAL Architect, IBM.

Thanks for looking into this patch Stewart

Stewart Smith [off-list ref] writes:

We're about to introduce an MPIPL reboot type (to take a firmware
assisted kdump style thing), and we maybe should have a reboot type to
force attempting a fast-reboot, and this makes me think if we should add
those in now?

I will probably let Vasant and others answer that.

If the reboot type isn't supported, what should be the behvaiour? Reboot
the default way or don't reboot at all?

Yes, I have addressed that in v3 of this patch at
http://patchwork.ozlabs.org/patch/967248/. In case the reboot type isnt
supported or if there is an error invoking it, the patch will revert
back to calling opal_cec_reboot().

-- 
Vaibhav Jain [off-list ref]
Linux Technology Center, IBM India Pvt. Ltd.