diff options
author | Greg Kurz <groug@kaod.org> | 2019-05-15 12:05:01 +0200 |
---|---|---|
committer | Michael Ellerman <mpe@ellerman.id.au> | 2019-06-02 19:39:36 +1000 |
commit | a3bf9fbdad600b1e4335dd90979f8d6072e4f602 (patch) | |
tree | 4bccd7a1334e2786ffe63e5546aad40fd38af802 /COPYING | |
parent | 02c5f5394918b9b47ff4357b1b18335768cd867d (diff) | |
download | linux-a3bf9fbdad600b1e4335dd90979f8d6072e4f602.tar.bz2 |
powerpc/pseries: Fix xive=off command line
On POWER9, if the hypervisor supports XIVE exploitation mode, the
guest OS will unconditionally requests for the XIVE interrupt mode
even if XIVE was deactivated with the kernel command line xive=off.
Later on, when the spapr XIVE init code handles xive=off, it disables
XIVE and tries to fall back on the legacy mode XICS.
This discrepency causes a kernel panic because the hypervisor is
configured to provide the XIVE interrupt mode to the guest :
kernel BUG at arch/powerpc/sysdev/xics/xics-common.c:135!
...
NIP xics_smp_probe+0x38/0x98
LR xics_smp_probe+0x2c/0x98
Call Trace:
xics_smp_probe+0x2c/0x98 (unreliable)
pSeries_smp_probe+0x40/0xa0
smp_prepare_cpus+0x62c/0x6ec
kernel_init_freeable+0x148/0x448
kernel_init+0x2c/0x148
ret_from_kernel_thread+0x5c/0x68
Look for xive=off during prom_init and don't ask for XIVE in this
case. One exception though: if the host only supports XIVE, we still
want to boot so we ignore xive=off.
Similarly, have the spapr XIVE init code to looking at the interrupt
mode negotiated during CAS, and ignore xive=off if the hypervisor only
supports XIVE.
Fixes: eac1e731b59e ("powerpc/xive: guest exploitation of the XIVE interrupt controller")
Cc: stable@vger.kernel.org # v4.20
Reported-by: Pavithra R. Prakash <pavrampu@in.ibm.com>
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Diffstat (limited to 'COPYING')
0 files changed, 0 insertions, 0 deletions