| Message ID | 20200605052345.1494-1-yuzenghui@huawei.com (mailing list archive) |
|---|---|
| State | Mainlined |
| Commit | 31dbb6b1d025506b3b8b8b74e9b697df47b9f696 |
| Headers | show |
| Series | irqchip/gic-v4.1: Use readx_poll_timeout_atomic() to fix sleep in atomic | expand |
Hi Marc, Sorry to ping you in the merge window, but ... On 2020/6/5 13:23, Zenghui Yu wrote: > readx_poll_timeout() can sleep if @sleep_us is specified by the caller, > and is therefore unsafe to be used inside the atomic context, which is > this case when we use it to poll the GICR_VPENDBASER.Dirty bit in > irq_set_vcpu_affinity() callback. this seems like an urgent thing to me. Without this patch, CPUs are easily to get stuck on my board with GICv4.1 enabled. So it'd be good if you can have a look and take this as a fix (if it is correct). Thanks, Zenghui > > Let's convert to its atomic version instead which helps to get the v4.1 > board back to life! > > Fixes: 96806229ca03 ("irqchip/gic-v4.1: Add support for VPENDBASER's Dirty+Valid signaling") > Signed-off-by: Zenghui Yu <yuzenghui@huawei.com> > --- > drivers/irqchip/irq-gic-v3-its.c | 8 ++++---- > 1 file changed, 4 insertions(+), 4 deletions(-) > > diff --git a/drivers/irqchip/irq-gic-v3-its.c b/drivers/irqchip/irq-gic-v3-its.c > index cd685f521c77..6a5a87fc4601 100644 > --- a/drivers/irqchip/irq-gic-v3-its.c > +++ b/drivers/irqchip/irq-gic-v3-its.c > @@ -3797,10 +3797,10 @@ static void its_wait_vpt_parse_complete(void) > if (!gic_rdists->has_vpend_valid_dirty) > return; > > - WARN_ON_ONCE(readq_relaxed_poll_timeout(vlpi_base + GICR_VPENDBASER, > - val, > - !(val & GICR_VPENDBASER_Dirty), > - 10, 500)); > + WARN_ON_ONCE(readq_relaxed_poll_timeout_atomic(vlpi_base + GICR_VPENDBASER, > + val, > + !(val & GICR_VPENDBASER_Dirty), > + 10, 500)); > } > > static void its_vpe_schedule(struct its_vpe *vpe) >
Hi Zenghui, On 2020-06-10 14:59, Zenghui Yu wrote: > Hi Marc, > > Sorry to ping you in the merge window, but ... > > On 2020/6/5 13:23, Zenghui Yu wrote: >> readx_poll_timeout() can sleep if @sleep_us is specified by the >> caller, >> and is therefore unsafe to be used inside the atomic context, which is >> this case when we use it to poll the GICR_VPENDBASER.Dirty bit in >> irq_set_vcpu_affinity() callback. > > this seems like an urgent thing to me. Without this patch, CPUs are > easily to get stuck on my board with GICv4.1 enabled. So it'd be good > if > you can have a look and take this as a fix (if it is correct). No worries. I've earmarked the patch for -rc1 already, just haven't got a chance to build the branch yet (a bit busy on the KVM side). I'll probably update the branch tonight or tomorrow. Thanks, M.
On Fri, 5 Jun 2020 13:23:45 +0800, Zenghui Yu wrote: > readx_poll_timeout() can sleep if @sleep_us is specified by the caller, > and is therefore unsafe to be used inside the atomic context, which is > this case when we use it to poll the GICR_VPENDBASER.Dirty bit in > irq_set_vcpu_affinity() callback. > > Let's convert to its atomic version instead which helps to get the v4.1 > board back to life! > > [...] Applied to irq/irqchip-next, thanks! [1/1] irqchip/gic-v4.1: Use readx_poll_timeout_atomic() to fix sleep in atomic commit: a87d4e00eacbc95b44466e3470529f4de49b450a Cheers, M.
diff --git a/drivers/irqchip/irq-gic-v3-its.c b/drivers/irqchip/irq-gic-v3-its.c index cd685f521c77..6a5a87fc4601 100644 --- a/drivers/irqchip/irq-gic-v3-its.c +++ b/drivers/irqchip/irq-gic-v3-its.c @@ -3797,10 +3797,10 @@ static void its_wait_vpt_parse_complete(void) if (!gic_rdists->has_vpend_valid_dirty) return; - WARN_ON_ONCE(readq_relaxed_poll_timeout(vlpi_base + GICR_VPENDBASER, - val, - !(val & GICR_VPENDBASER_Dirty), - 10, 500)); + WARN_ON_ONCE(readq_relaxed_poll_timeout_atomic(vlpi_base + GICR_VPENDBASER, + val, + !(val & GICR_VPENDBASER_Dirty), + 10, 500)); } static void its_vpe_schedule(struct its_vpe *vpe)
readx_poll_timeout() can sleep if @sleep_us is specified by the caller, and is therefore unsafe to be used inside the atomic context, which is this case when we use it to poll the GICR_VPENDBASER.Dirty bit in irq_set_vcpu_affinity() callback. Let's convert to its atomic version instead which helps to get the v4.1 board back to life! Fixes: 96806229ca03 ("irqchip/gic-v4.1: Add support for VPENDBASER's Dirty+Valid signaling") Signed-off-by: Zenghui Yu <yuzenghui@huawei.com> --- drivers/irqchip/irq-gic-v3-its.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)