On Tue, Sep 01, 2020 at 11:37:30PM +0300, Dmitry Osipenko wrote: > The mapping operations of the Tegra SMMU driver are subjected to a race > condition issues because SMMU Address Space isn't allocated and freed > atomically, while it should be. This patch makes the mapping operations > atomic, it fixes an accidentally released Host1x Address Space problem > which happens while running multiple graphics tests in parallel on > Tegra30, i.e. by having multiple threads racing with each other in the > Host1x's submission and completion code paths, performing IOVA mappings > and unmappings in parallel. > > Signed-off-by: Dmitry Osipenko <digetx@xxxxxxxxx> > --- > > Changelog: > > v5: - Replaced GFP_NOWAIT check with __GFP_ATOMIC to fix "sleep in > atomic context" warnings, NOWAIT != ATOMIC. > > v4: - Returned to use spinlock, but now using a smarter allocation > logic that performs allocation in a sleeping context whenever > possible. > > - Removed the stable tag because patch isn't portable as-is > since the arguments of map/unmap() callbacks changed recently. > Perhaps we could just ignore older kernels for now. It will be > possible to fix older kernels with a custom patch if will be needed. > > v3: - No changes. Resending for visibility. > > drivers/iommu/tegra-smmu.c | 95 +++++++++++++++++++++++++++++++++----- > 1 file changed, 84 insertions(+), 11 deletions(-) Seems to work fine. Tested on Jetson TX1 with display and GPU, which are the primary users of the SMMU. Tested-by: Thierry Reding <treding@xxxxxxxxxx> Acked-by: Thierry Reding <treding@xxxxxxxxxx>
Attachment:
signature.asc
Description: PGP signature
