On 24/05/2022 16:06, Greg Kroah-Hartman wrote:
On Tue, May 24, 2022 at 03:55:58PM +0100, Jon Hunter wrote:
On 24/05/2022 13:09, Greg Kroah-Hartman wrote:
...
I am seeing a boot regression on tegra124-jetson-tk1 and reverting the above commit is fixing the problem. This also appears to impact linux-4.14.y, 4.19.y and 5.4.y.
Test results for stable-v4.9: 8 builds: 8 pass, 0 fail 18 boots: 16 pass, 2 fail 18 tests: 18 pass, 0 fail
Linux version: 4.9.316-rc1-gbe4ec3e3faa1 Boards tested: tegra124-jetson-tk1, tegra20-ventana, tegra210-p2371-2180, tegra30-cardhu-a04
Boot failures: tegra124-jetson-tk1
Odd. This is also in 5.10.y, right? No issues there? Are we missing something?
Actually, the more I look at this, the more I see various intermittent reports with this and it is also impacting the mainline.
The problem is that the commit in question is causing a ton of messages to be printed a boot and this sometimes is causing the boot test to fail because the boot is taking too long. The console shows ...
[ 1233.327547] CPU0: Spectre BHB: using loop workaround [ 1233.327795] CPU1: Spectre BHB: using loop workaround [ 1233.328270] CPU1: Spectre BHB: using loop workaround [ 1233.328700] CPU1: Spectre BHB: using loop workaround [ 1233.355477] CPU2: Spectre BHB: using loop workaround ** 7 printk messages dropped ** [ 1233.366271] CPU0: Spectre BHB: using loop workaround [ 1233.366580] CPU0: Spectre BHB: using loop workaround [ 1233.366815] CPU1: Spectre BHB: using loop workaround [ 1233.405475] CPU1: Spectre BHB: using loop workaround [ 1233.405874] CPU0: Spectre BHB: using loop workaround [ 1233.406041] CPU1: Spectre BHB: using loop workaround ** 1 printk messages dropped **
There is a similar report of this [0] and I believe that we need a similar fix for the above prints as well. I have reported this to Ard [1]. So I am not sure that these Spectre BHB patches are quite ready for stable.
These patches are quite small, and just enable it for this known-broken cpu type.
If there is an issue enabling it for this cpu type, then we can work on that upstream, but there shouldn't be a reason to prevent this from being merged now, especially given that it is supposed to be fixing a known issue.
Yes understand. I have been doing some more testing and with v4.9, this is triggering a boot timeout 100% of the time. So with 20 boots, all 20 timeout. Note the timeout is 2 mins. With v4.14, I saw only 5 out of 20 timeouts and so it would seem that v4.9 is slower to boot in general. I think that the more recent kernels show intermittent timeouts.
We have some verbose logging enabled on this platform, which until now has not been a problem, but I will disable this and that should resolve this for now.
Cheers Jon