| Message ID | 20170601154353.9208-1-hegdevasant@linux.vnet.ibm.com |
|---|---|
| State | Accepted |
| Headers | show |
On Thu, 2017-06-01 at 21:13 +0530, Vasant Hegde wrote: > We use irq for reading input from console, but not in output path. > Hence do not enable input irq in write path. > > Fixes : 583c8203 (fsp/console: Allocate irq for each hvc console) > CC: Sam Mendoza-Jonas <sam@mendozajonas.com> > Signed-off-by: Vasant Hegde <hegdevasant@linux.vnet.ibm.com> As encountered by Stewart, fixes a soft-lockup on ZZ with a 4.12.1 kernel. Acked-By: Samuel Mendoza-Jonas <sam@mendozajonas.com> > --- > hw/fsp/fsp-console.c | 3 --- > 1 file changed, 3 deletions(-) > > diff --git a/hw/fsp/fsp-console.c b/hw/fsp/fsp-console.c > index 8eaf7b8..fd67b20 100644 > --- a/hw/fsp/fsp-console.c > +++ b/hw/fsp/fsp-console.c > @@ -193,7 +193,6 @@ static size_t fsp_write_vserial(struct fsp_serial *fs, const char *buf, > #ifndef DISABLE_CON_PENDING_EVT > opal_update_pending_evt(OPAL_EVENT_CONSOLE_OUTPUT, > OPAL_EVENT_CONSOLE_OUTPUT); > - opal_update_pending_evt(fs->irq, fs->irq); > #endif > return len; > } > @@ -746,12 +745,10 @@ void fsp_console_poll(void *data __unused) > if (!fs->open) > continue; > if (sb->next_out == sb->next_in) { > - opal_update_pending_evt(fs->irq, 0); > continue; > } > if (fs->log_port) { > flush_console(); > - opal_update_pending_evt(fs->irq, 0); > } else { > #ifdef OPAL_DEBUG_CONSOLE_POLL > if (debug < 5) {
Vasant Hegde <hegdevasant@linux.vnet.ibm.com> writes: > We use irq for reading input from console, but not in output path. > Hence do not enable input irq in write path. > > Fixes : 583c8203 (fsp/console: Allocate irq for each hvc console) > CC: Sam Mendoza-Jonas <sam@mendozajonas.com> > Signed-off-by: Vasant Hegde <hegdevasant@linux.vnet.ibm.com> Thanks, merged to master as of 72aa080d0970f7757ccf4f3b00096f2f8bbcafde. I also now have a test case that seems independent of skiboot version, where you can get a soft lockup in kopald on recent kernels.. This seems to be limited to when you exit petitboot and go for a fast reboot *really* soon after booting. I can usually reproduce within 5 fast-reboots, but sometimes takes as much as 10. I'm pretty convinced that this is a kernel issue.. or at least not a *regression* on the skiboot side since our service pack stream. So, I think we should get to the bottom of that before I necessarily backport this fix to 5.4.x? Ideally, it's probably about time some time is spent on getting that giant test matrix for these things automated....
diff --git a/hw/fsp/fsp-console.c b/hw/fsp/fsp-console.c index 8eaf7b8..fd67b20 100644 --- a/hw/fsp/fsp-console.c +++ b/hw/fsp/fsp-console.c @@ -193,7 +193,6 @@ static size_t fsp_write_vserial(struct fsp_serial *fs, const char *buf, #ifndef DISABLE_CON_PENDING_EVT opal_update_pending_evt(OPAL_EVENT_CONSOLE_OUTPUT, OPAL_EVENT_CONSOLE_OUTPUT); - opal_update_pending_evt(fs->irq, fs->irq); #endif return len; } @@ -746,12 +745,10 @@ void fsp_console_poll(void *data __unused) if (!fs->open) continue; if (sb->next_out == sb->next_in) { - opal_update_pending_evt(fs->irq, 0); continue; } if (fs->log_port) { flush_console(); - opal_update_pending_evt(fs->irq, 0); } else { #ifdef OPAL_DEBUG_CONSOLE_POLL if (debug < 5) {
We use irq for reading input from console, but not in output path. Hence do not enable input irq in write path. Fixes : 583c8203 (fsp/console: Allocate irq for each hvc console) CC: Sam Mendoza-Jonas <sam@mendozajonas.com> Signed-off-by: Vasant Hegde <hegdevasant@linux.vnet.ibm.com> --- hw/fsp/fsp-console.c | 3 --- 1 file changed, 3 deletions(-)