[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <871qi7qupw.ffs@tglx>
Date: Mon, 19 Jun 2023 18:31:55 +0200
From: Thomas Gleixner <tglx@...utronix.de>
To: Dheeraj Kumar Srivastava <dheerajkumar.srivastava@....com>,
linux-kernel@...r.kernel.org, mingo@...hat.com, bp@...en8.de,
dave.hansen@...ux.intel.com, x86@...nel.org
Cc: hpa@...or.com, gorcunov@...nvz.org, suresh.b.siddha@...el.com,
Dheeraj Kumar Srivastava <dheerajkumar.srivastava@....com>,
Kishon Vijay Abraham I <kvijayab@....com>,
Vasant Hegde <vasant.hegde@....com>
Subject: Re: [PATCH v1] x86/apic: Fix kernel panic when "intremap=off" and
"x2apic_phys" are set
On Sat, Jun 17 2023 at 02:52, Dheeraj Kumar Srivastava wrote:
> x2APIC mode requires "Interrupt Remapping" to be enabled and the
> "physical x2apic" driver can be used only when x2APIC mode is enabled.
> However when "intremap=off" and "x2apic_phys" kernel command line
> parameters are passed, "physical x2apic" driver is being used even when
> x2APIC mode is disabled ("intremap=off" disables x2APIC mode).
> This results in the below kernel panic:
>
> unchecked MSR access error: RDMSR from 0x80f at rIP: 0xffffffff87eab4ec
> (native_read_msr+0xc/0x40)
> Call Trace:
> <TASK>
> native_apic_msr_read+0x1f/0x30
> setup_local_APIC+0x4e/0x380
> ? zen_untrain_ret+0x1/0x1
> ? enable_IR_x2apic+0xe8/0x250
> apic_intr_mode_init+0x4c/0x120
> x86_late_time_init+0x28/0x40
> start_kernel+0x626/0xa80
> x86_64_start_reservations+0x1c/0x30
> x86_64_start_kernel+0xbf/0x110
> secondary_startup_64_no_verify+0x10b/0x10b
> </TASK>
>
> This is due to an incorrect conditional check in x2apic_phys_probe().
> Fix it here by returning 0 when "x2apic_mode" is not set in
> x2apic_phys_probe(). This now prevents default_setup_apic_routing() from
> selecting "physical x2apic" driver.
>
> Fixes: 9ebd680bd029 ("x86, apic: Use probe routines to simplify apic selection")
> Reviewed-by: Kishon Vijay Abraham I <kvijayab@....com>
> Reviewed-by: Vasant Hegde <vasant.hegde@....com>
> Signed-off-by: Dheeraj Kumar Srivastava <dheerajkumar.srivastava@....com>
Reviewed-by: Thomas Gleixner <tglx@...utronix.de>
Powered by blists - more mailing lists