David Gibson <da...@gibson.dropbear.id.au> writes:
> On Tue, Jul 16, 2019 at 01:13:52PM +0100, Alex Bennée wrote: >> The opcode decode tables aren't really part of the CPUPPCState but an >> internal implementation detail for the translator. This can cause >> problems with memcpy in cpu_copy as any table created during >> ppc_cpu_realize get written over causing a memory leak. To avoid this >> move the tables into PowerPCCPU which is better suited to hold >> internal implementation details. >> >> Attempts to fix: https://bugs.launchpad.net/qemu/+bug/1836558 >> Cc: 1836...@bugs.launchpad.net >> Signed-off-by: Alex Bennée <alex.ben...@linaro.org> > > I've applied this now to ppc-for-4.2. If there's an argument for > including it in 4.1 during hard freeze, you'll need to spell it out > for me. Well without: Subject: [RFC PATCH for 4.1] linux-user: unparent CPU object before unref Date: Tue, 16 Jul 2019 15:01:33 +0100 Message-Id: <20190716140133.8578-1-alex.ben...@linaro.org> it doesn't matter as we never attempt to free the memory once a thread is destroyed. This causes all linux-user guests that create and destroy threads quickly to slowly leak memory. However due to the dynamic opcode table ppc/ppc64-linux-user guests leak a lot faster than most, in the order of ~50k each time a thread is created and destroyed. However I'm happy to defer to you as the maintainer :-) -- Alex Bennée
