Thread (7 messages) flat view 7 messages, 4 authors, 2021-08-06

Re: [PATCH] powerpc/xive: Do not skip CPU-less nodes when creating the IPIs

From: Cédric Le Goater <clg@kaod.org>
Date: 2021-08-02 13:35:13
Also in: stable

On 8/2/21 8:37 AM, Michael Ellerman wrote:
Cédric Le Goater [off-list ref] writes:
quoted
On PowerVM, CPU-less nodes can be populated with hot-plugged CPUs at
runtime. Today, the IPI is not created for such nodes, and hot-plugged
CPUs use a bogus IPI, which leads to soft lockups.

We could create the node IPI on demand but it is a bit complex because
this code would be called under bringup_up() and some IRQ locking is
being done. The simplest solution is to create the IPIs for all nodes
at startup.

Fixes: 7dcc37b3eff9 ("powerpc/xive: Map one IPI interrupt per node")
Cc: stable@vger.kernel.org # v5.13
Reported-by: Geetika Moolchandani <redacted>
Cc: Srikar Dronamraju <redacted>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
---

This patch breaks old versions of irqbalance (<= v1.4). Possible nodes
are collected from /sys/devices/system/node/ but CPU-less nodes are
not listed there. When interrupts are scanned, the link representing
the node structure is NULL and segfault occurs.
Breaking userspace is usually frowned upon, even if it is irqbalance.

If CPU-less nodes appeared in /sys/devices/system/node would that fix
it? Could we do that or is that not possible for other reasons?
quoted
Version 1.7 seems immune. 
Which was released in August 2020.

Looks like some distros still ship 1.6, I take it you're not sure if
that is broken or not.
I did a bisect on irqbalance and the "bad" commit was introduced between 
version 1.7 and version 1.8 :

  commit 31dea01f3a47 ("Also fetch node info for non-PCI devices") 
  https://github.com/Irqbalance/irqbalance/commit/31dea01f3a47aa6374560638486879e5129f9c94

which was backported on RHEL 8 in RPM irqbalance-1.4.0-6.el8.

Any distro using irqbalance <= 1.7 without the patch above is fine. 

Since irqbalance handled cleanly irqs referencing offline nodes before 
this patch, I am inclined to think that the irqbalance fix is incomplete. 
Unfortunately, the commit log lacks some context on the non-PCI devices. 

Thanks,

C. 
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help