| Message ID | alpine.DEB.2.10.1605191726030.15478@laptop (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 05/19/2016 03:31 PM, Scot Doyle wrote: > Two systems are locking on boot [1] because ops->cur_blink_jiffies > is set to zero from vc->vc_cur_blink_ms. > > Ignore such invalid intervals and log a warning. > > [1] https://bugs.launchpad.net/bugs/1574814 > > Suggested-by: David Daney <david.daney@cavium.com> > Signed-off-by: Scot Doyle <lkml14@scotdoyle.com> > Cc: <stable@vger.kernel.org> [v4.2] This seems better. I didn't test it, but... Acked-by: David Daney <david.daney@cavium.com> > --- > drivers/video/console/fbcon.c | 14 ++++++++++++-- > 1 file changed, 12 insertions(+), 2 deletions(-) > > diff --git a/drivers/video/console/fbcon.c b/drivers/video/console/fbcon.c > index 6e92917..fad5b89 100644 > --- a/drivers/video/console/fbcon.c > +++ b/drivers/video/console/fbcon.c > @@ -1095,7 +1095,13 @@ static void fbcon_init(struct vc_data *vc, int init) > con_copy_unimap(vc, svc); > > ops = info->fbcon_par; > - ops->cur_blink_jiffies = msecs_to_jiffies(vc->vc_cur_blink_ms); > + > + if (vc->vc_cur_blink_ms >= 50) > + ops->cur_blink_jiffies = > + msecs_to_jiffies(vc->vc_cur_blink_ms); > + else > + WARN_ONCE(1, "blink interval < 50 ms"); > + > p->con_rotate = initial_rotation; > set_blitting_type(vc, info); > > @@ -1309,7 +1315,11 @@ static void fbcon_cursor(struct vc_data *vc, int mode) > int y; > int c = scr_readw((u16 *) vc->vc_pos); > > - ops->cur_blink_jiffies = msecs_to_jiffies(vc->vc_cur_blink_ms); > + if (vc->vc_cur_blink_ms >= 50) > + ops->cur_blink_jiffies = > + msecs_to_jiffies(vc->vc_cur_blink_ms); > + else > + WARN_ONCE(1, "blink interval < 50 ms"); > > if (fbcon_is_inactive(vc, info) || vc->vc_deccm != 1) > return; > -- To unsubscribe from this list: send the line "unsubscribe linux-fbdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Scot, > Two systems are locking on boot [1] because ops->cur_blink_jiffies > is set to zero from vc->vc_cur_blink_ms. > > Ignore such invalid intervals and log a warning. This prevents a lockup on AST BMC machines, but (as expected) generates a warning against the fbcon driver, which is a significantly better result. Tested-by: Jeremy Kerr <jk@ozlabs.org> [now to sort out the issue in the ast driver...] Cheers, Jeremy -- To unsubscribe from this list: send the line "unsubscribe linux-fbdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, May 20, 2016 at 6:31 AM, Scot Doyle <lkml14@scotdoyle.com> wrote: > Two systems are locking on boot [1] because ops->cur_blink_jiffies > is set to zero from vc->vc_cur_blink_ms. > > Ignore such invalid intervals and log a warning. > > [1] https://bugs.launchpad.net/bugs/1574814 > > Suggested-by: David Daney <david.daney@cavium.com> > Signed-off-by: Scot Doyle <lkml14@scotdoyle.com> > Cc: <stable@vger.kernel.org> [v4.2] Not sure this one is needed for stable because it justs dumps a warning, and not set a valid period to ops->cur_blink_jiffies. So I guess other fix patch is still required for the soft lockup issue, right? Thanks, > --- > drivers/video/console/fbcon.c | 14 ++++++++++++-- > 1 file changed, 12 insertions(+), 2 deletions(-) > > diff --git a/drivers/video/console/fbcon.c b/drivers/video/console/fbcon.c > index 6e92917..fad5b89 100644 > --- a/drivers/video/console/fbcon.c > +++ b/drivers/video/console/fbcon.c > @@ -1095,7 +1095,13 @@ static void fbcon_init(struct vc_data *vc, int init) > con_copy_unimap(vc, svc); > > ops = info->fbcon_par; > - ops->cur_blink_jiffies = msecs_to_jiffies(vc->vc_cur_blink_ms); > + > + if (vc->vc_cur_blink_ms >= 50) > + ops->cur_blink_jiffies = > + msecs_to_jiffies(vc->vc_cur_blink_ms); > + else > + WARN_ONCE(1, "blink interval < 50 ms"); > + > p->con_rotate = initial_rotation; > set_blitting_type(vc, info); > > @@ -1309,7 +1315,11 @@ static void fbcon_cursor(struct vc_data *vc, int mode) > int y; > int c = scr_readw((u16 *) vc->vc_pos); > > - ops->cur_blink_jiffies = msecs_to_jiffies(vc->vc_cur_blink_ms); > + if (vc->vc_cur_blink_ms >= 50) > + ops->cur_blink_jiffies = > + msecs_to_jiffies(vc->vc_cur_blink_ms); > + else > + WARN_ONCE(1, "blink interval < 50 ms"); > > if (fbcon_is_inactive(vc, info) || vc->vc_deccm != 1) > return; > -- > 2.1.4 > -- To unsubscribe from this list: send the line "unsubscribe linux-fbdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Ming, > Not sure this one is needed for stable because it justs dumps > a warning, and not set a valid period to ops->cur_blink_jiffies. > > So I guess other fix patch is still required for the soft lockup > issue, right? The main thing is that we don't set cur_blink_jiffies to the < 50ms value. As far as I can tell, it means we'll still get the original default, set in fbcon_startup(): ops->cur_blink_jiffies = HZ / 5; And so don't end up spinning on the timer expiry. Cheers, Jeremy -- To unsubscribe from this list: send the line "unsubscribe linux-fbdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, May 20, 2016 at 10:26 AM, Jeremy Kerr <jk@ozlabs.org> wrote: > Hi Ming, > >> Not sure this one is needed for stable because it justs dumps >> a warning, and not set a valid period to ops->cur_blink_jiffies. >> >> So I guess other fix patch is still required for the soft lockup >> issue, right? > > The main thing is that we don't set cur_blink_jiffies to the < 50ms > value. As far as I can tell, it means we'll still get the original > default, set in fbcon_startup(): > > ops->cur_blink_jiffies = HZ / 5; > > And so don't end up spinning on the timer expiry. Jeremy, your theory is correct, thanks for your clarification! And my test just shows that this patch does fix the soft lockup too, so Tested-by: Ming Lei <ming.lei@canonical.com> Then looks there are two fix patches acked & tested: - the patch in this thread - another one "[PATCH] tty: vt: Fix soft lockup in fbcon cursor blink timer." https://lkml.org/lkml/2016/5/17/455 So which one will be pushed to linus? Thanks, Ming > > Cheers, > > > Jeremy -- To unsubscribe from this list: send the line "unsubscribe linux-fbdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Ming, >Then looks there are two fix patches acked & tested: > > - the patch in this thread > - another one "[PATCH] tty: vt: Fix soft lockup in fbcon cursor >blink timer." > https://lkml.org/lkml/2016/5/17/455 > >So which one will be pushed to linus? Not that it's my call, but we may want both; the first as a safety measure to prevent an invalid cur_blink_jiffies ever being set, and the second one to actually fix the initialisation of vc_cur_blink_ms (and address the warning introduced by the first). I guess we could just go with the latter for stable... Cheers, Jeremy -- To unsubscribe from this list: send the line "unsubscribe linux-fbdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, 20 May 2016, Jeremy Kerr wrote: > Hi Ming, > > >Then looks there are two fix patches acked & tested: > > > > - the patch in this thread > > - another one "[PATCH] tty: vt: Fix soft lockup in fbcon cursor > >blink timer." > > https://lkml.org/lkml/2016/5/17/455 > > > >So which one will be pushed to linus? > > Not that it's my call, but we may want both; the first as a safety > measure to prevent an invalid cur_blink_jiffies ever being set, and the > second one to actually fix the initialisation of vc_cur_blink_ms (and > address the warning introduced by the first). Tomi / Greg, I'd suggest - applying "tty: vt: Fix soft lockup in fbcon cursor blink timer." to 4.7 and stable[4.2] - applying "fbcon: warn on invalid cursor blink intervals" to 4.7 - ignoring "fbcon: use default if cursor blink interval is not valid" Note: the patches don't depend on each other > I guess we could just go with the latter for stable... > > Cheers, > > Jeremy -- To unsubscribe from this list: send the line "unsubscribe linux-fbdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, 20 May 2016, Scot Doyle wrote: > On Fri, 20 May 2016, Jeremy Kerr wrote: > > Hi Ming, > > > > >Then looks there are two fix patches acked & tested: > > > > > > - the patch in this thread > > > - another one "[PATCH] tty: vt: Fix soft lockup in fbcon cursor > > >blink timer." > > > https://lkml.org/lkml/2016/5/17/455 > > > > > >So which one will be pushed to linus? > > > > Not that it's my call, but we may want both; the first as a safety > > measure to prevent an invalid cur_blink_jiffies ever being set, and the > > second one to actually fix the initialisation of vc_cur_blink_ms (and > > address the warning introduced by the first). > > Tomi / Greg, > > I'd suggest > - applying "tty: vt: Fix soft lockup in fbcon cursor blink timer." to 4.7 and stable[4.2] > - applying "fbcon: warn on invalid cursor blink intervals" to 4.7 > - ignoring "fbcon: use default if cursor blink interval is not valid" > > Note: the patches don't depend on each other "tty: vt: Fix soft lockup..." should be applied first in order to avoid unnecessary reports due to the log warning in "fbcon: warn on invalid..." > > > I guess we could just go with the latter for stable... > > > > Cheers, > > > > Jeremy -- To unsubscribe from this list: send the line "unsubscribe linux-fbdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Thu, 19 May 2016, Scot Doyle wrote: > Two systems are locking on boot [1] because ops->cur_blink_jiffies > is set to zero from vc->vc_cur_blink_ms. > > Ignore such invalid intervals and log a warning. > > [1] https://bugs.launchpad.net/bugs/1574814 > > Suggested-by: David Daney <david.daney@cavium.com> > Signed-off-by: Scot Doyle <lkml14@scotdoyle.com> > Cc: <stable@vger.kernel.org> [v4.2] FWIW: Tested-by: Henrique de Moraes Holschuh <hmh@hmh.eng.br> on top of 4.4.11. And nothing caused it to issue warnings here, so far (with the other recommended patch applied first). > --- > drivers/video/console/fbcon.c | 14 ++++++++++++-- > 1 file changed, 12 insertions(+), 2 deletions(-) > > diff --git a/drivers/video/console/fbcon.c b/drivers/video/console/fbcon.c > index 6e92917..fad5b89 100644 > --- a/drivers/video/console/fbcon.c > +++ b/drivers/video/console/fbcon.c > @@ -1095,7 +1095,13 @@ static void fbcon_init(struct vc_data *vc, int init) > con_copy_unimap(vc, svc); > > ops = info->fbcon_par; > - ops->cur_blink_jiffies = msecs_to_jiffies(vc->vc_cur_blink_ms); > + > + if (vc->vc_cur_blink_ms >= 50) > + ops->cur_blink_jiffies = > + msecs_to_jiffies(vc->vc_cur_blink_ms); > + else > + WARN_ONCE(1, "blink interval < 50 ms"); > + > p->con_rotate = initial_rotation; > set_blitting_type(vc, info); > > @@ -1309,7 +1315,11 @@ static void fbcon_cursor(struct vc_data *vc, int mode) > int y; > int c = scr_readw((u16 *) vc->vc_pos); > > - ops->cur_blink_jiffies = msecs_to_jiffies(vc->vc_cur_blink_ms); > + if (vc->vc_cur_blink_ms >= 50) > + ops->cur_blink_jiffies = > + msecs_to_jiffies(vc->vc_cur_blink_ms); > + else > + WARN_ONCE(1, "blink interval < 50 ms"); > > if (fbcon_is_inactive(vc, info) || vc->vc_deccm != 1) > return;
On Fri, 20 May 2016, Scot Doyle wrote: > On Fri, 20 May 2016, Jeremy Kerr wrote: > > >Then looks there are two fix patches acked & tested: > > > > > > - the patch in this thread > > > - another one "[PATCH] tty: vt: Fix soft lockup in fbcon cursor > > >blink timer." > > > https://lkml.org/lkml/2016/5/17/455 > > > > > >So which one will be pushed to linus? > > > > Not that it's my call, but we may want both; the first as a safety > > measure to prevent an invalid cur_blink_jiffies ever being set, and the > > second one to actually fix the initialisation of vc_cur_blink_ms (and > > address the warning introduced by the first). > > Tomi / Greg, > > I'd suggest > - applying "tty: vt: Fix soft lockup in fbcon cursor blink timer." to 4.7 and stable[4.2] > - applying "fbcon: warn on invalid cursor blink intervals" to 4.7 > - ignoring "fbcon: use default if cursor blink interval is not valid" > > Note: the patches don't depend on each other I applied both recommended patches on top of 4.4.11 for testing, and they made things a lot better here. I suggest the second patch should be backported to stable too, might as well fix this thing for good *and keep the door closed*.
> -----Original Message----- > From: Henrique de Moraes Holschuh [mailto:hmh@hmh.eng.br] > Sent: Saturday, May 28, 2016 4:49 AM > To: Scot Doyle <lkml14@scotdoyle.com> > Cc: Tomi Valkeinen <tomi.valkeinen@ti.com>; Jean-Christophe Plagniol- > Villard <plagnioj@jcrosoft.com>; Greg Kroah-Hartman > <gregkh@linuxfoundation.org>; Jeremy Kerr <jk@ozlabs.org>; Ming Lei > <ming.lei@canonical.com>; Daney, David <David.Daney@cavium.com>; > Dann Frazier <dann.frazier@canonical.com>; Peter Hurley > <peter@hurleysoftware.com>; Pavel Machek <pavel@ucw.cz>; Jonathan Liu > <net147@gmail.com>; Alistair Popple <alistair@popple.id.au>; Jean-Philippe > Brucker <jean-philippe.brucker@arm.com>; Chintakuntla, Radha > <Radha.Chintakuntla@cavium.com>; Jiri Slaby <jslaby@suse.com>; David > Airlie <airlied@linux.ie>; David Daney <ddaney.cavm@gmail.com>; dri- > devel@lists.freedesktop.org; linux-fbdev@vger.kernel.org; Linux Kernel > Mailing List <linux-kernel@vger.kernel.org>; stable > <stable@vger.kernel.org> > Subject: Re: [PATCH] fbcon: warn on invalid cursor blink intervals > > On Fri, 20 May 2016, Scot Doyle wrote: > > On Fri, 20 May 2016, Jeremy Kerr wrote: > > > >Then looks there are two fix patches acked & tested: > > > > > > > > - the patch in this thread > > > > - another one "[PATCH] tty: vt: Fix soft lockup in fbcon cursor > > > >blink timer." > > > > https://lkml.org/lkml/2016/5/17/455 > > > > > > > >So which one will be pushed to linus? > > > > > > Not that it's my call, but we may want both; the first as a safety > > > measure to prevent an invalid cur_blink_jiffies ever being set, and the > > > second one to actually fix the initialisation of vc_cur_blink_ms (and > > > address the warning introduced by the first). > > > > Tomi / Greg, > > > > I'd suggest > > - applying "tty: vt: Fix soft lockup in fbcon cursor blink timer." to 4.7 and > stable[4.2] > > - applying "fbcon: warn on invalid cursor blink intervals" to 4.7 > > - ignoring "fbcon: use default if cursor blink interval is not valid" > > > > Note: the patches don't depend on each other > > I applied both recommended patches on top of 4.4.11 for testing, and they > made things a lot better here. > > I suggest the second patch should be backported to stable too, might as well > fix this thing for good *and keep the door closed*. Is this patch available on some tree so that I can point to ? And hope it will make it to linux-next soon ? > > -- > "One disk to rule them all, One disk to find them. One disk to bring > them all and in the darkness grind them. In the Land of Redmond > where the shadows lie." -- The Silicon Valley Tarot > Henrique Holschuh -- To unsubscribe from this list: send the line "unsubscribe linux-fbdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/video/console/fbcon.c b/drivers/video/console/fbcon.c index 6e92917..fad5b89 100644 --- a/drivers/video/console/fbcon.c +++ b/drivers/video/console/fbcon.c @@ -1095,7 +1095,13 @@ static void fbcon_init(struct vc_data *vc, int init) con_copy_unimap(vc, svc); ops = info->fbcon_par; - ops->cur_blink_jiffies = msecs_to_jiffies(vc->vc_cur_blink_ms); + + if (vc->vc_cur_blink_ms >= 50) + ops->cur_blink_jiffies = + msecs_to_jiffies(vc->vc_cur_blink_ms); + else + WARN_ONCE(1, "blink interval < 50 ms"); + p->con_rotate = initial_rotation; set_blitting_type(vc, info); @@ -1309,7 +1315,11 @@ static void fbcon_cursor(struct vc_data *vc, int mode) int y; int c = scr_readw((u16 *) vc->vc_pos); - ops->cur_blink_jiffies = msecs_to_jiffies(vc->vc_cur_blink_ms); + if (vc->vc_cur_blink_ms >= 50) + ops->cur_blink_jiffies = + msecs_to_jiffies(vc->vc_cur_blink_ms); + else + WARN_ONCE(1, "blink interval < 50 ms"); if (fbcon_is_inactive(vc, info) || vc->vc_deccm != 1) return;
Two systems are locking on boot [1] because ops->cur_blink_jiffies is set to zero from vc->vc_cur_blink_ms. Ignore such invalid intervals and log a warning. [1] https://bugs.launchpad.net/bugs/1574814 Suggested-by: David Daney <david.daney@cavium.com> Signed-off-by: Scot Doyle <lkml14@scotdoyle.com> Cc: <stable@vger.kernel.org> [v4.2] --- drivers/video/console/fbcon.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-)