Date:   Wed, 12 Jan 2022 16:14:19 +0000
From:   Sean Christopherson <seanjc@google.com>
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     kernel test robot <oliver.sang@intel.com>,
        Josh Poimboeuf <jpoimboe@redhat.com>,
        LKML <linux-kernel@vger.kernel.org>, x86@kernel.org,
        lkp@lists.01.org, lkp@intel.com
Subject: Re: [x86/entry_32]  aa93e2ad74:
 BUG:soft_lockup-CPU##stuck_for#s![systemd-logind:#]
Message-ID: <Yd7+W1gFbkBs5Lkd@google.com>
References: <20220106083523.GB32167@xsang-OptiPlex-9020>
 <Yd1l0gInc4zRcnt/@hirez.programming.kicks-ass.net>
 <Yd4u2rVVSdpEpwwM@google.com>
 <Yd6ssarpDknV1r9z@hirez.programming.kicks-ass.net>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <Yd6ssarpDknV1r9z@hirez.programming.kicks-ass.net>
Precedence: bulk

On Wed, Jan 12, 2022, Peter Zijlstra wrote:
> On Wed, Jan 12, 2022 at 01:28:58AM +0000, Sean Christopherson wrote:
> > The issue is that PARANOID_EXIT_TO_KERNEL_MODE in the handle_exception_return
> > path overwrites the entry stack data with the task stack data, restoring the "bad"
> > segment value.
> 
> Oh gawd... that's terrible, and yes, that now makes perfect sense.
> 
> However did you find that?

printf and running under QEMU, which has a interactive "monitor" that lets you
read/write guest memory and can also do VA=>PA translations.  Code inspection
once I realized the value on the stack was being restored between the exception
fixup and the POP.