[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-Id: <20200518234245.200672-1-leobras.c@gmail.com>
Date: Mon, 18 May 2020 20:42:43 -0300
From: Leonardo Bras <leobras.c@...il.com>
To: Michael Ellerman <mpe@...erman.id.au>,
Benjamin Herrenschmidt <benh@...nel.crashing.org>,
Paul Mackerras <paulus@...ba.org>,
Allison Randal <allison@...utok.net>,
Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
Thomas Gleixner <tglx@...utronix.de>,
Nicholas Piggin <npiggin@...il.com>,
Leonardo Bras <leobras.c@...il.com>,
Nathan Lynch <nathanl@...ux.ibm.com>,
"Gautham R. Shenoy" <ego@...ux.vnet.ibm.com>,
Nadav Amit <namit@...are.com>
Cc: linuxppc-dev@...ts.ozlabs.org, linux-kernel@...r.kernel.org
Subject: [PATCH v6 0/2] Implement reentrant rtas call
Patch 2 implement rtas_call_reentrant() for reentrant rtas-calls:
"ibm,int-on", "ibm,int-off",ibm,get-xive" and "ibm,set-xive",
according to LoPAPR Version 1.1 (March 24, 2016).
For that, it's necessary that every call uses a different
rtas buffer (rtas_args). Paul Mackerras suggested using the PACA
structure for creating a per-cpu buffer for these calls.
Patch 1 was necessary to make PACA have a 'struct rtas_args' member.
Reentrant rtas calls can be useful to avoid deadlocks in crashing,
where rtas-calls are needed, but some other thread crashed holding
the rtas.lock.
This is a backtrace of a deadlock from a kdump testing environment:
#0 arch_spin_lock
#1 lock_rtas ()
#2 rtas_call (token=8204, nargs=1, nret=1, outputs=0x0)
#3 ics_rtas_mask_real_irq (hw_irq=4100)
#4 machine_kexec_mask_interrupts
#5 default_machine_crash_shutdown
#6 machine_crash_shutdown
#7 __crash_kexec
#8 crash_kexec
#9 oops_end
Signed-off-by: Leonardo Bras <leobras.c@...il.com>
Special thanks to Nick Piggin, who have been helping me a lot with
this series!
---
Changes since v5:
- Renames new paca member from reentrant_args to rtas_args_reentrant
- Compile out rtas_args_reentrant if CONFIG_PPC_RTAS=n
- new_rtas_args() is skipped (returns NULL) if CPU_FTR_HVMODE
Changes since v4:
- Insted of having the full buffer on PACA, adds only a pointer and
allocate it during allocate_paca(), making sure it's in a memory
range available for RTAS (32-bit). (Thanks Nick Piggin!)
Changes since v3:
- Adds protection from preemption and interruption
Changes since v2:
- Fixed build failure from ppc64e, by including spinlock_types.h on
rtas-types.h
- Improved commit messages
Changes since v1:
- Moved buffer from stack to PACA (as suggested by Paul Mackerras)
- Added missing output bits
- Improve documentation following kernel-doc format (as suggested by
Nathan Lynch)
Leonardo Bras (2):
powerpc/rtas: Move type/struct definitions from rtas.h into
rtas-types.h
powerpc/rtas: Implement reentrant rtas call
arch/powerpc/include/asm/paca.h | 2 +
arch/powerpc/include/asm/rtas-types.h | 126 ++++++++++++++++++++++++++
arch/powerpc/include/asm/rtas.h | 119 +-----------------------
arch/powerpc/kernel/rtas.c | 42 +++++++++
arch/powerpc/sysdev/xics/ics-rtas.c | 22 ++---
5 files changed, 183 insertions(+), 128 deletions(-)
create mode 100644 arch/powerpc/include/asm/rtas-types.h
--
2.25.4
Powered by blists - more mailing lists