| Message ID | 1458659073-13553-1-git-send-email-tomeu.vizoso@collabora.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Tue, Mar 22, 2016 at 04:04:33PM +0100, Tomeu Vizoso wrote: > Move the logging of timeouts when stalling the MMU to > rk_iommu_attach_device, as it's expected that sometimes the MMU won't > get stalled when detaching a device, and it's not a real problem that > would need to be communicated to the user. > > Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com> Heiko, any opinion on this patch?
Am Dienstag, 5. April 2016, 16:57:26 schrieb Joerg Roedel: > On Tue, Mar 22, 2016 at 04:04:33PM +0100, Tomeu Vizoso wrote: > > Move the logging of timeouts when stalling the MMU to > > rk_iommu_attach_device, as it's expected that sometimes the MMU won't > > get stalled when detaching a device, and it's not a real problem that > > would need to be communicated to the user. > > > > Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com> > > Heiko, any opinion on this patch? I don't know enough about iommus (and the rockchip implementation) to judge if this is correct. The reasoning sound right though and it seems to fix an issue for Tomeu. One complaint would be that I'd expect the behaviour to match between the enable and disable functions. Meaning I'd want the error message from rk_iommu_disable_stall to also move rk_iommu_attach_device. Heiko
Am Dienstag, 5. April 2016, 09:48:15 schrieb Heiko Stuebner: > Am Dienstag, 5. April 2016, 16:57:26 schrieb Joerg Roedel: > > On Tue, Mar 22, 2016 at 04:04:33PM +0100, Tomeu Vizoso wrote: > > > Move the logging of timeouts when stalling the MMU to > > > rk_iommu_attach_device, as it's expected that sometimes the MMU won't > > > get stalled when detaching a device, and it's not a real problem that > > > would need to be communicated to the user. > > > > > > Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com> > > > > Heiko, any opinion on this patch? > > I don't know enough about iommus (and the rockchip implementation) to > judge if this is correct. The reasoning sound right though and it seems > to fix an issue for Tomeu. > > One complaint would be that I'd expect the behaviour to match between the > enable and disable functions. Meaning I'd want the error message from > rk_iommu_disable_stall to also move rk_iommu_attach_device. hmm, or does John's patch maybe tie into this problem? Aka, does the iommu simply does not detect when it's already stalled and emits the timeout then? Tomeu, in which cases do you see the excess error message?
diff --git a/drivers/iommu/rockchip-iommu.c b/drivers/iommu/rockchip-iommu.c index 0253ab35c33b..65325b6742e6 100644 --- a/drivers/iommu/rockchip-iommu.c +++ b/drivers/iommu/rockchip-iommu.c @@ -346,13 +346,7 @@ static int rk_iommu_enable_stall(struct rk_iommu *iommu) rk_iommu_command(iommu, RK_MMU_CMD_ENABLE_STALL); - ret = rk_wait_for(rk_iommu_is_stall_active(iommu), 1); - if (ret) - for (i = 0; i < iommu->num_mmu; i++) - dev_err(iommu->dev, "Enable stall request timed out, status: %#08x\n", - rk_iommu_read(iommu->bases[i], RK_MMU_STATUS)); - - return ret; + return rk_wait_for(rk_iommu_is_stall_active(iommu), 1); } static int rk_iommu_disable_stall(struct rk_iommu *iommu) @@ -798,8 +792,12 @@ static int rk_iommu_attach_device(struct iommu_domain *domain, return 0; ret = rk_iommu_enable_stall(iommu); - if (ret) + if (ret) { + for (i = 0; i < iommu->num_mmu; i++) + dev_err(iommu->dev, "Enable stall request timed out, status: %#08x\n", + rk_iommu_read(iommu->bases[i], RK_MMU_STATUS)); return ret; + } ret = rk_iommu_force_reset(iommu); if (ret)
Move the logging of timeouts when stalling the MMU to rk_iommu_attach_device, as it's expected that sometimes the MMU won't get stalled when detaching a device, and it's not a real problem that would need to be communicated to the user. Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com> --- drivers/iommu/rockchip-iommu.c | 14 ++++++-------- 1 file changed, 6 insertions(+), 8 deletions(-)