On Wed 2018-06-27 16:08:14, Petr Mladek wrote:
The commit 719f6a7040f1bdaf96 ("printk: Use the main logbuf in NMI when logbuf_lock is available") brought back the possible deadlocks in printk() and NMI.
This is rework of the proposed fix, see https://lkml.kernel.org/r/20180606111557.xzs6l3lkvg7lq3ts@pathway.suse.cz
I realized that we could rather easily move the check to vprintk_func() and still avoid any race. I believe that this is a win-win solution.
Changes against v1:
- Move the check from vprintk_emit() to vprintk_func()
- More straightforward commit message
- Fix build with CONFIG_PRINTK_NMI disabled
Petr Mladek (3): printk: Split the code for storing a message into the log buffer printk: Create helper function to queue deferred console handling printk/nmi: Prevent deadlock when accessing the main log buffer in NMI
include/linux/printk.h | 4 ++++ kernel/printk/internal.h | 9 ++++++- kernel/printk/printk.c | 57 +++++++++++++++++++++++++++----------------- kernel/printk/printk_safe.c | 58 +++++++++++++++++++++++++++++---------------- kernel/trace/trace.c | 4 +++- lib/nmi_backtrace.c | 3 --- 6 files changed, 87 insertions(+), 48 deletions(-)
JFYI, I have renamed the new function defer_console() to defer_console_output() and pushed everything to printk.git, branch for-4.19-nmi.
Best Regards, Petr