* Linux Kernel soft hangs if 8259 PIC is not configured by BIOS: was (RE: [edk2-devel] [RFC][PATCH v1 0/1] PcAtChipsetPkg: Remove framework modules) @ 2019-05-28 16:36 Kirkendall, Garrett 2019-05-28 17:06 ` Michael D Kinney 2019-05-28 17:17 ` Leif Lindholm 0 siblings, 2 replies; 4+ messages in thread From: Kirkendall, Garrett @ 2019-05-28 16:36 UTC (permalink / raw) To: devel@edk2.groups.io, hao.a.wu@intel.com Cc: Ray Ni, Andrew Fish, Laszlo Ersek, Leif Lindholm, Michael D Kinney Since this change will be removing the 8259 PIC driver, I'll mention a quirk I ran into in the Linux Kernel up through at least 5.0 (and I am sure later, I just didn't verify): Does anyone have pull in the Linux Kernel world to get them to take this seriously? Hopefully someday soon end users will let legacy devices, like the PIC, die a long overdue and graceful death. I sent the following to the responsible Linux Kernel developers and quickly got back a "working as designed" response: ************************* I am trying to boot a UEFI BIOS with minimal legacy hardware support. The Linux kernel soft hangs when the PIC is not configured by the BIOS because it is using IOAPIC. Hopefully, this provides enough information. Observed under Ubuntu Server Linux 18.04 LTS with kernel 4.15.0, and with kernel compiled from source tag v5.0 Where it hangs: Soft hang occurs in calibrate_APIC_clock(): https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/apic/apic.c#L805 specific location of soft hang waiting for interrupts: https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/apic/apic.c#L854 How it gets to the hang: If 8259A PIC is not configured before kernel is launched, HPET IRQ 0 registration fails because probe_8259A returns PIC as not available and therefore interrupt descriptors 0-15 are not allocated. This happens when BIOS does not configure 8259A PIC because it uses IOAPIC. This sequence prevents allocating interrupts 0-15 unless PIC is configured before kernel starts. legacy_pic.init = init_8259A is not called before early_irq_init(): early_irq_init(): https://github.com/torvalds/linux/blob/v5.0/init/main.c#L642 initcnt = arch_probe_nr_irqs(): https://github.com/torvalds/linux/blob/v5.0/kernel/irq/irqdesc.c#L512 return legacy_pic->probe(): https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/apic/vector.c#L656 default_legacy_pic.probe: https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/i8259.c#L418 probe_8259A(): https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/i8259.c#L301 interrupt 0-15 descriptors not allocated and prevents IOAPIC interrupts 0-15: https://github.com/torvalds/linux/blob/v5.0/kernel/irq/irqdesc.c#L525 In this call is where init_8259A is called and PIC is initialized. However, interrupt descriptors for 0-15 were not allocated in early_irq_init() sequence, so descriptors are not available later. init_IRQ() : https://github.com/torvalds/linux/blob/v5.0/init/main.c#L643 This sequence tries to register the HPET to irq0, but irq0 descriptor is not allocated by early_irq_init() sequence: late_time_init(): https://github.com/torvalds/linux/blob/v5.0/init/main.c#L703 late_time_init = x86_late_time_init: https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/time.c#L107 x86_init.timers.timer_init(): https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/time.c#L92 x86_init.timers.timer_init = hpet_time_init: https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/x86_init.c#L75 setup_default_timer_irq(): https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/time.c#L83 if (setup_irq(0, &irq0)): https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/time.c#L78 This gets called at some point after the above sequences, I couldn't track it all the way back to main.c easily: x86_init.timers.setup_percpu_clockev = setup_boot_APIC_clock: https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/x86_init.c#L74 https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/apic/apic.c#L961 *** soft hang in calibrate_APIC_clock(): https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/apic/apic.c#L854 ************************* GARRETT KIRKENDALL SMTS Firmware Engineer | CTE 7171 Southwest Parkway, Austin, TX 78735 USA AMD facebook | amd.com -----Original Message----- From: devel@edk2.groups.io <devel@edk2.groups.io> On Behalf Of Wu, Hao A via Groups.Io Sent: Monday, May 27, 2019 1:37 AM To: devel@edk2.groups.io Cc: Hao A Wu <hao.a.wu@intel.com>; Ray Ni <ray.ni@intel.com>; Andrew Fish <afish@apple.com>; Laszlo Ersek <lersek@redhat.com>; Leif Lindholm <leif.lindholm@linaro.org>; Michael D Kinney <michael.d.kinney@intel.com> Subject: [edk2-devel] [RFC][PATCH v1 0/1] PcAtChipsetPkg: Remove framework modules [CAUTION: External Email] ''' Please note that this patch will be hold until all the below requirements are met: A. edk2-stable201905 is created; B. OvmfPkg has drop its usage of the legacy ISA stack (which includes the IsaAcpiDxe driver); C. UefiPayloadPkg has been updated to drop its consume to the 8259InterruptControllerDxe driver. ''' This patch itself will not be sent to the mailing list and is only available at: https://github.com/hwu25/edk2/tree/delete_framework (https://github.com/hwu25/edk2/commit/84e58f89f6d03f9cc3399cced9d5d5529e06a416) Below modules will be removed from PcAtChipsetPkg: * PcAtChipsetPkg/8259InterruptControllerDxe/8259.inf * PcAtChipsetPkg/8254TimerDxe/8254Timer.inf * PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.inf They are considered legacy framework components and will no longer be used after the removal of IntelFramework[Module]Pkg. Also, the unused (after the modules being removed) PCDs will be deleted in package level DEC/UNI files. Cc: Ray Ni <ray.ni@intel.com> Cc: Andrew Fish <afish@apple.com> Cc: Laszlo Ersek <lersek@redhat.com> Cc: Leif Lindholm <leif.lindholm@linaro.org> Cc: Michael D Kinney <michael.d.kinney@intel.com> Hao A Wu (1): PcAtChipsetPkg: Remove framework modules PcAtChipsetPkg/PcAtChipsetPkg.dec | 58 -- PcAtChipsetPkg/PcAtChipsetPkg.dsc | 5 +- PcAtChipsetPkg/8254TimerDxe/8254Timer.inf | 42 -- PcAtChipsetPkg/8259InterruptControllerDxe/8259.inf | 46 -- PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.inf | 50 -- PcAtChipsetPkg/8254TimerDxe/Timer.h | 185 ------ PcAtChipsetPkg/8259InterruptControllerDxe/8259.h | 220 ------- PcAtChipsetPkg/IsaAcpiDxe/PcatIsaAcpi.h | 269 --------- PcAtChipsetPkg/8254TimerDxe/Timer.c | 401 ------------- PcAtChipsetPkg/8259InterruptControllerDxe/8259.c | 622 -------------------- PcAtChipsetPkg/IsaAcpiDxe/ComponentName.c | 301 ---------- PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.c | 353 ----------- PcAtChipsetPkg/IsaAcpiDxe/PcatIsaAcpi.c | 386 ------------ PcAtChipsetPkg/8254TimerDxe/Timer.uni | 16 - PcAtChipsetPkg/8254TimerDxe/TimerExtra.uni | 14 - PcAtChipsetPkg/8259InterruptControllerDxe/Legacy8259.uni | 16 - PcAtChipsetPkg/8259InterruptControllerDxe/Legacy8259Extra.uni | 14 - PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.uni | 16 - PcAtChipsetPkg/IsaAcpiDxe/IsaAcpiExtra.uni | 14 - PcAtChipsetPkg/PcAtChipsetPkg.uni | 52 +- 20 files changed, 2 insertions(+), 3078 deletions(-) delete mode 100644 PcAtChipsetPkg/8254TimerDxe/8254Timer.inf delete mode 100644 PcAtChipsetPkg/8259InterruptControllerDxe/8259.inf delete mode 100644 PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.inf delete mode 100644 PcAtChipsetPkg/8254TimerDxe/Timer.h delete mode 100644 PcAtChipsetPkg/8259InterruptControllerDxe/8259.h delete mode 100644 PcAtChipsetPkg/IsaAcpiDxe/PcatIsaAcpi.h delete mode 100644 PcAtChipsetPkg/8254TimerDxe/Timer.c delete mode 100644 PcAtChipsetPkg/8259InterruptControllerDxe/8259.c delete mode 100644 PcAtChipsetPkg/IsaAcpiDxe/ComponentName.c delete mode 100644 PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.c delete mode 100644 PcAtChipsetPkg/IsaAcpiDxe/PcatIsaAcpi.c delete mode 100644 PcAtChipsetPkg/8254TimerDxe/Timer.uni delete mode 100644 PcAtChipsetPkg/8254TimerDxe/TimerExtra.uni delete mode 100644 PcAtChipsetPkg/8259InterruptControllerDxe/Legacy8259.uni delete mode 100644 PcAtChipsetPkg/8259InterruptControllerDxe/Legacy8259Extra.uni delete mode 100644 PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.uni delete mode 100644 PcAtChipsetPkg/IsaAcpiDxe/IsaAcpiExtra.uni -- 2.12.0.windows.1 ^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: Linux Kernel soft hangs if 8259 PIC is not configured by BIOS: was (RE: [edk2-devel] [RFC][PATCH v1 0/1] PcAtChipsetPkg: Remove framework modules) 2019-05-28 16:36 Linux Kernel soft hangs if 8259 PIC is not configured by BIOS: was (RE: [edk2-devel] [RFC][PATCH v1 0/1] PcAtChipsetPkg: Remove framework modules) Kirkendall, Garrett @ 2019-05-28 17:06 ` Michael D Kinney 2019-05-28 17:31 ` Kirkendall, Garrett 2019-05-28 17:17 ` Leif Lindholm 1 sibling, 1 reply; 4+ messages in thread From: Michael D Kinney @ 2019-05-28 17:06 UTC (permalink / raw) To: Kirkendall, Garrett, devel@edk2.groups.io, Wu, Hao A, Kinney, Michael D Cc: Ni, Ray, Andrew Fish, Laszlo Ersek, Leif Lindholm Hi Garrett, Two thoughts on this issue: 1) Just because 8259 is not used by UEFI FW does not mean the 8259 HW should not be initialized into an OS compatible state. This init code could be part of the chipset/si init code without requiring the 8259 driver/protocol. 2) The HPET driver in PcAtChipsetPkg can be configured to use IOAPIC or MSI delivery for interrupts. Is there a way for Linux to also use MSI delivery instead of IOAPIC delivery? Thanks, Mike > -----Original Message----- > From: Kirkendall, Garrett > [mailto:Garrett.Kirkendall@amd.com] > Sent: Tuesday, May 28, 2019 9:36 AM > To: devel@edk2.groups.io; Wu, Hao A > <hao.a.wu@intel.com> > Cc: Ni, Ray <ray.ni@intel.com>; Andrew Fish > <afish@apple.com>; Laszlo Ersek <lersek@redhat.com>; > Leif Lindholm <leif.lindholm@linaro.org>; Kinney, > Michael D <michael.d.kinney@intel.com> > Subject: Linux Kernel soft hangs if 8259 PIC is not > configured by BIOS: was (RE: [edk2-devel] [RFC][PATCH > v1 0/1] PcAtChipsetPkg: Remove framework modules) > > Since this change will be removing the 8259 PIC driver, > I'll mention a quirk I ran into in the Linux Kernel up > through at least 5.0 (and I am sure later, I just > didn't verify): > > Does anyone have pull in the Linux Kernel world to get > them to take this seriously? Hopefully someday soon > end users will let legacy devices, like the PIC, die a > long overdue and graceful death. > > I sent the following to the responsible Linux Kernel > developers and quickly got back a "working as designed" > response: > > ************************* > I am trying to boot a UEFI BIOS with minimal legacy > hardware support. The Linux kernel soft hangs when the > PIC is not configured by the BIOS because it is using > IOAPIC. Hopefully, this provides enough information. > > Observed under Ubuntu Server Linux 18.04 LTS with > kernel 4.15.0, and with kernel compiled from source tag > v5.0 > > Where it hangs: > Soft hang occurs in calibrate_APIC_clock(): > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/apic/apic.c#L805 > specific location of soft hang waiting for interrupts: > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/apic/apic.c#L854 > > > How it gets to the hang: > If 8259A PIC is not configured before kernel is > launched, HPET IRQ 0 registration fails because > probe_8259A returns PIC as not available and therefore > interrupt descriptors 0-15 are not allocated. This > happens when BIOS does not configure 8259A PIC because > it uses IOAPIC. > > This sequence prevents allocating interrupts 0-15 > unless PIC is configured before kernel starts. > legacy_pic.init = init_8259A is not called before > early_irq_init(): > early_irq_init(): > https://github.com/torvalds/linux/blob/v5.0/init/main.c > #L642 > initcnt = arch_probe_nr_irqs(): > https://github.com/torvalds/linux/blob/v5.0/kernel/irq/ > irqdesc.c#L512 > return legacy_pic->probe(): > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/apic/vector.c#L656 > default_legacy_pic.probe: > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/i8259.c#L418 > probe_8259A(): > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/i8259.c#L301 > interrupt 0-15 descriptors not allocated and > prevents IOAPIC interrupts 0-15: > https://github.com/torvalds/linux/blob/v5.0/kernel/irq/ > irqdesc.c#L525 > > In this call is where init_8259A is called and PIC is > initialized. However, interrupt descriptors for 0-15 > were not allocated in early_irq_init() sequence, so > descriptors are not available later. > init_IRQ() : > https://github.com/torvalds/linux/blob/v5.0/init/main.c > #L643 > > This sequence tries to register the HPET to irq0, but > irq0 descriptor is not allocated by early_irq_init() > sequence: > late_time_init(): > https://github.com/torvalds/linux/blob/v5.0/init/main.c > #L703 > late_time_init = x86_late_time_init: > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/time.c#L107 > x86_init.timers.timer_init(): > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/time.c#L92 > x86_init.timers.timer_init = hpet_time_init: > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/x86_init.c#L75 > setup_default_timer_irq(): > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/time.c#L83 > if (setup_irq(0, &irq0)): > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/time.c#L78 > > This gets called at some point after the above > sequences, I couldn't track it all the way back to > main.c easily: > x86_init.timers.setup_percpu_clockev = > setup_boot_APIC_clock: > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/x86_init.c#L74 > > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/apic/apic.c#L961 > *** soft hang in calibrate_APIC_clock(): > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/apic/apic.c#L854 > > ************************* > > GARRETT KIRKENDALL > SMTS Firmware Engineer | CTE > 7171 Southwest Parkway, Austin, TX 78735 USA > AMD facebook | amd.com > > -----Original Message----- > From: devel@edk2.groups.io <devel@edk2.groups.io> On > Behalf Of Wu, Hao A via Groups.Io > Sent: Monday, May 27, 2019 1:37 AM > To: devel@edk2.groups.io > Cc: Hao A Wu <hao.a.wu@intel.com>; Ray Ni > <ray.ni@intel.com>; Andrew Fish <afish@apple.com>; > Laszlo Ersek <lersek@redhat.com>; Leif Lindholm > <leif.lindholm@linaro.org>; Michael D Kinney > <michael.d.kinney@intel.com> > Subject: [edk2-devel] [RFC][PATCH v1 0/1] > PcAtChipsetPkg: Remove framework modules > > [CAUTION: External Email] > > ''' > Please note that this patch will be hold until all the > below requirements are met: > > A. edk2-stable201905 is created; > B. OvmfPkg has drop its usage of the legacy ISA stack > (which includes > the IsaAcpiDxe driver); > C. UefiPayloadPkg has been updated to drop its consume > to the > 8259InterruptControllerDxe driver. > ''' > > This patch itself will not be sent to the mailing list > and is only available at: > https://github.com/hwu25/edk2/tree/delete_framework > (https://github.com/hwu25/edk2/commit/84e58f89f6d03f9cc > 3399cced9d5d5529e06a416) > > > Below modules will be removed from PcAtChipsetPkg: > * PcAtChipsetPkg/8259InterruptControllerDxe/8259.inf > * PcAtChipsetPkg/8254TimerDxe/8254Timer.inf > * PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.inf > > They are considered legacy framework components and > will no longer be used after the removal of > IntelFramework[Module]Pkg. > > Also, the unused (after the modules being removed) PCDs > will be deleted in package level DEC/UNI files. > > Cc: Ray Ni <ray.ni@intel.com> > Cc: Andrew Fish <afish@apple.com> > Cc: Laszlo Ersek <lersek@redhat.com> > Cc: Leif Lindholm <leif.lindholm@linaro.org> > Cc: Michael D Kinney <michael.d.kinney@intel.com> > > > Hao A Wu (1): > PcAtChipsetPkg: Remove framework modules > > PcAtChipsetPkg/PcAtChipsetPkg.dec > | 58 -- > PcAtChipsetPkg/PcAtChipsetPkg.dsc > | 5 +- > PcAtChipsetPkg/8254TimerDxe/8254Timer.inf > | 42 -- > PcAtChipsetPkg/8259InterruptControllerDxe/8259.inf > | 46 -- > PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.inf > | 50 -- > PcAtChipsetPkg/8254TimerDxe/Timer.h > | 185 ------ > PcAtChipsetPkg/8259InterruptControllerDxe/8259.h > | 220 ------- > PcAtChipsetPkg/IsaAcpiDxe/PcatIsaAcpi.h > | 269 --------- > PcAtChipsetPkg/8254TimerDxe/Timer.c > | 401 ------------- > PcAtChipsetPkg/8259InterruptControllerDxe/8259.c > | 622 -------------------- > PcAtChipsetPkg/IsaAcpiDxe/ComponentName.c > | 301 ---------- > PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.c > | 353 ----------- > PcAtChipsetPkg/IsaAcpiDxe/PcatIsaAcpi.c > | 386 ------------ > PcAtChipsetPkg/8254TimerDxe/Timer.uni > | 16 - > PcAtChipsetPkg/8254TimerDxe/TimerExtra.uni > | 14 - > > PcAtChipsetPkg/8259InterruptControllerDxe/Legacy8259.un > i | 16 - > > PcAtChipsetPkg/8259InterruptControllerDxe/Legacy8259Ext > ra.uni | 14 - > PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.uni > | 16 - > PcAtChipsetPkg/IsaAcpiDxe/IsaAcpiExtra.uni > | 14 - > PcAtChipsetPkg/PcAtChipsetPkg.uni > | 52 +- > 20 files changed, 2 insertions(+), 3078 deletions(-) > delete mode 100644 > PcAtChipsetPkg/8254TimerDxe/8254Timer.inf > delete mode 100644 > PcAtChipsetPkg/8259InterruptControllerDxe/8259.inf > delete mode 100644 > PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.inf > delete mode 100644 PcAtChipsetPkg/8254TimerDxe/Timer.h > delete mode 100644 > PcAtChipsetPkg/8259InterruptControllerDxe/8259.h > delete mode 100644 > PcAtChipsetPkg/IsaAcpiDxe/PcatIsaAcpi.h > delete mode 100644 PcAtChipsetPkg/8254TimerDxe/Timer.c > delete mode 100644 > PcAtChipsetPkg/8259InterruptControllerDxe/8259.c > delete mode 100644 > PcAtChipsetPkg/IsaAcpiDxe/ComponentName.c > delete mode 100644 PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.c > delete mode 100644 > PcAtChipsetPkg/IsaAcpiDxe/PcatIsaAcpi.c > delete mode 100644 > PcAtChipsetPkg/8254TimerDxe/Timer.uni > delete mode 100644 > PcAtChipsetPkg/8254TimerDxe/TimerExtra.uni > delete mode 100644 > PcAtChipsetPkg/8259InterruptControllerDxe/Legacy8259.un > i > delete mode 100644 > PcAtChipsetPkg/8259InterruptControllerDxe/Legacy8259Ext > ra.uni > delete mode 100644 > PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.uni > delete mode 100644 > PcAtChipsetPkg/IsaAcpiDxe/IsaAcpiExtra.uni > > -- > 2.12.0.windows.1 > > > ^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: Linux Kernel soft hangs if 8259 PIC is not configured by BIOS: was (RE: [edk2-devel] [RFC][PATCH v1 0/1] PcAtChipsetPkg: Remove framework modules) 2019-05-28 17:06 ` Michael D Kinney @ 2019-05-28 17:31 ` Kirkendall, Garrett 0 siblings, 0 replies; 4+ messages in thread From: Kirkendall, Garrett @ 2019-05-28 17:31 UTC (permalink / raw) To: Kinney, Michael D, devel@edk2.groups.io, Wu, Hao A Cc: Ni, Ray, Andrew Fish, Laszlo Ersek, Leif Lindholm See below -----Original Message----- From: Kinney, Michael D <michael.d.kinney@intel.com> Sent: Tuesday, May 28, 2019 12:07 PM To: Kirkendall, Garrett <Garrett.Kirkendall@amd.com>; devel@edk2.groups.io; Wu, Hao A <hao.a.wu@intel.com>; Kinney, Michael D <michael.d.kinney@intel.com> Cc: Ni, Ray <ray.ni@intel.com>; Andrew Fish <afish@apple.com>; Laszlo Ersek <lersek@redhat.com>; Leif Lindholm <leif.lindholm@linaro.org> Subject: RE: Linux Kernel soft hangs if 8259 PIC is not configured by BIOS: was (RE: [edk2-devel] [RFC][PATCH v1 0/1] PcAtChipsetPkg: Remove framework modules) [CAUTION: External Email] Hi Garrett, Two thoughts on this issue: 1) Just because 8259 is not used by UEFI FW does not mean the 8259 HW should not be initialized into an OS compatible state. This init code could be part of the chipset/si init code without requiring the 8259 driver/protocol. [Garrett:] The Linux Kernel only allocates "descriptors"/structures for Interrupts 0-15, PIC or IOAPIC mode, if a legacy PIC is available and responds the expected way. The Kernel will initialize the PIC, but not until after this initial probe of the PIC. If the PIC simply did not exist in the hardware, these interrupts would not be allowed/allocated even for an IOAPIC. I was hoping to not need to do anything with the legacy PIC since I had no plans to use it for this UEFI implementation. 2) The HPET driver in PcAtChipsetPkg can be configured to use IOAPIC or MSI delivery for interrupts. Is there a way for Linux to also use MSI delivery instead of IOAPIC delivery? [Garrett:] This is a very early timer initialization sequence and the PIC/IOAPIC will be fully initialized later. Since the PIC didn't respond correctly, I don't think interrupts 0-15 would be allowed later. I'm not sure if the kernel allows MSI for the HPET when everything is configured, I didn't go slogging through to figure out. Thanks, Mike > -----Original Message----- > From: Kirkendall, Garrett > [mailto:Garrett.Kirkendall@amd.com] > Sent: Tuesday, May 28, 2019 9:36 AM > To: devel@edk2.groups.io; Wu, Hao A > <hao.a.wu@intel.com> > Cc: Ni, Ray <ray.ni@intel.com>; Andrew Fish <afish@apple.com>; Laszlo > Ersek <lersek@redhat.com>; Leif Lindholm <leif.lindholm@linaro.org>; > Kinney, Michael D <michael.d.kinney@intel.com> > Subject: Linux Kernel soft hangs if 8259 PIC is not configured by > BIOS: was (RE: [edk2-devel] [RFC][PATCH > v1 0/1] PcAtChipsetPkg: Remove framework modules) > > Since this change will be removing the 8259 PIC driver, I'll mention a > quirk I ran into in the Linux Kernel up through at least 5.0 (and I am > sure later, I just didn't verify): > > Does anyone have pull in the Linux Kernel world to get them to take > this seriously? Hopefully someday soon end users will let legacy > devices, like the PIC, die a long overdue and graceful death. > > I sent the following to the responsible Linux Kernel developers and > quickly got back a "working as designed" > response: > > ************************* > I am trying to boot a UEFI BIOS with minimal legacy hardware support. > The Linux kernel soft hangs when the PIC is not configured by the BIOS > because it is using IOAPIC. Hopefully, this provides enough > information. > > Observed under Ubuntu Server Linux 18.04 LTS with kernel 4.15.0, and > with kernel compiled from source tag > v5.0 > > Where it hangs: > Soft hang occurs in calibrate_APIC_clock(): > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/apic/apic.c#L805 > specific location of soft hang waiting for interrupts: > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/apic/apic.c#L854 > > > How it gets to the hang: > If 8259A PIC is not configured before kernel is launched, HPET IRQ 0 > registration fails because probe_8259A returns PIC as not available > and therefore interrupt descriptors 0-15 are not allocated. This > happens when BIOS does not configure 8259A PIC because it uses IOAPIC. > > This sequence prevents allocating interrupts 0-15 unless PIC is > configured before kernel starts. > legacy_pic.init = init_8259A is not called before > early_irq_init(): > early_irq_init(): > https://github.com/torvalds/linux/blob/v5.0/init/main.c > #L642 > initcnt = arch_probe_nr_irqs(): > https://github.com/torvalds/linux/blob/v5.0/kernel/irq/ > irqdesc.c#L512 > return legacy_pic->probe(): > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/apic/vector.c#L656 > default_legacy_pic.probe: > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/i8259.c#L418 > probe_8259A(): > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/i8259.c#L301 > interrupt 0-15 descriptors not allocated and prevents IOAPIC > interrupts 0-15: > https://github.com/torvalds/linux/blob/v5.0/kernel/irq/ > irqdesc.c#L525 > > In this call is where init_8259A is called and PIC is initialized. > However, interrupt descriptors for 0-15 were not allocated in > early_irq_init() sequence, so descriptors are not available later. > init_IRQ() : > https://github.com/torvalds/linux/blob/v5.0/init/main.c > #L643 > > This sequence tries to register the HPET to irq0, but > irq0 descriptor is not allocated by early_irq_init() > sequence: > late_time_init(): > https://github.com/torvalds/linux/blob/v5.0/init/main.c > #L703 > late_time_init = x86_late_time_init: > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/time.c#L107 > x86_init.timers.timer_init(): > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/time.c#L92 > x86_init.timers.timer_init = hpet_time_init: > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/x86_init.c#L75 > setup_default_timer_irq(): > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/time.c#L83 > if (setup_irq(0, &irq0)): > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/time.c#L78 > > This gets called at some point after the above sequences, I couldn't > track it all the way back to main.c easily: > x86_init.timers.setup_percpu_clockev = > setup_boot_APIC_clock: > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/x86_init.c#L74 > > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/apic/apic.c#L961 > *** soft hang in calibrate_APIC_clock(): > https://github.com/torvalds/linux/blob/v5.0/arch/x86/ke > rnel/apic/apic.c#L854 > > ************************* > > GARRETT KIRKENDALL > SMTS Firmware Engineer | CTE > 7171 Southwest Parkway, Austin, TX 78735 USA > AMD facebook | amd.com > > -----Original Message----- > From: devel@edk2.groups.io <devel@edk2.groups.io> On Behalf Of Wu, Hao > A via Groups.Io > Sent: Monday, May 27, 2019 1:37 AM > To: devel@edk2.groups.io > Cc: Hao A Wu <hao.a.wu@intel.com>; Ray Ni <ray.ni@intel.com>; Andrew > Fish <afish@apple.com>; Laszlo Ersek <lersek@redhat.com>; Leif > Lindholm <leif.lindholm@linaro.org>; Michael D Kinney > <michael.d.kinney@intel.com> > Subject: [edk2-devel] [RFC][PATCH v1 0/1] > PcAtChipsetPkg: Remove framework modules > > [CAUTION: External Email] > > ''' > Please note that this patch will be hold until all the below > requirements are met: > > A. edk2-stable201905 is created; > B. OvmfPkg has drop its usage of the legacy ISA stack (which includes > the IsaAcpiDxe driver); > C. UefiPayloadPkg has been updated to drop its consume to the > 8259InterruptControllerDxe driver. > ''' > > This patch itself will not be sent to the mailing list and is only > available at: > https://github.com/hwu25/edk2/tree/delete_framework > (https://github.com/hwu25/edk2/commit/84e58f89f6d03f9cc > 3399cced9d5d5529e06a416) > > > Below modules will be removed from PcAtChipsetPkg: > * PcAtChipsetPkg/8259InterruptControllerDxe/8259.inf > * PcAtChipsetPkg/8254TimerDxe/8254Timer.inf > * PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.inf > > They are considered legacy framework components and will no longer be > used after the removal of IntelFramework[Module]Pkg. > > Also, the unused (after the modules being removed) PCDs will be > deleted in package level DEC/UNI files. > > Cc: Ray Ni <ray.ni@intel.com> > Cc: Andrew Fish <afish@apple.com> > Cc: Laszlo Ersek <lersek@redhat.com> > Cc: Leif Lindholm <leif.lindholm@linaro.org> > Cc: Michael D Kinney <michael.d.kinney@intel.com> > > > Hao A Wu (1): > PcAtChipsetPkg: Remove framework modules > > PcAtChipsetPkg/PcAtChipsetPkg.dec > | 58 -- > PcAtChipsetPkg/PcAtChipsetPkg.dsc > | 5 +- > PcAtChipsetPkg/8254TimerDxe/8254Timer.inf > | 42 -- > PcAtChipsetPkg/8259InterruptControllerDxe/8259.inf > | 46 -- > PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.inf > | 50 -- > PcAtChipsetPkg/8254TimerDxe/Timer.h > | 185 ------ > PcAtChipsetPkg/8259InterruptControllerDxe/8259.h > | 220 ------- > PcAtChipsetPkg/IsaAcpiDxe/PcatIsaAcpi.h > | 269 --------- > PcAtChipsetPkg/8254TimerDxe/Timer.c > | 401 ------------- > PcAtChipsetPkg/8259InterruptControllerDxe/8259.c > | 622 -------------------- > PcAtChipsetPkg/IsaAcpiDxe/ComponentName.c > | 301 ---------- > PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.c > | 353 ----------- > PcAtChipsetPkg/IsaAcpiDxe/PcatIsaAcpi.c > | 386 ------------ > PcAtChipsetPkg/8254TimerDxe/Timer.uni > | 16 - > PcAtChipsetPkg/8254TimerDxe/TimerExtra.uni > | 14 - > > PcAtChipsetPkg/8259InterruptControllerDxe/Legacy8259.un > i | 16 - > > PcAtChipsetPkg/8259InterruptControllerDxe/Legacy8259Ext > ra.uni | 14 - > PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.uni > | 16 - > PcAtChipsetPkg/IsaAcpiDxe/IsaAcpiExtra.uni > | 14 - > PcAtChipsetPkg/PcAtChipsetPkg.uni > | 52 +- > 20 files changed, 2 insertions(+), 3078 deletions(-) delete mode > 100644 PcAtChipsetPkg/8254TimerDxe/8254Timer.inf > delete mode 100644 > PcAtChipsetPkg/8259InterruptControllerDxe/8259.inf > delete mode 100644 > PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.inf > delete mode 100644 PcAtChipsetPkg/8254TimerDxe/Timer.h > delete mode 100644 > PcAtChipsetPkg/8259InterruptControllerDxe/8259.h > delete mode 100644 > PcAtChipsetPkg/IsaAcpiDxe/PcatIsaAcpi.h > delete mode 100644 PcAtChipsetPkg/8254TimerDxe/Timer.c > delete mode 100644 > PcAtChipsetPkg/8259InterruptControllerDxe/8259.c > delete mode 100644 > PcAtChipsetPkg/IsaAcpiDxe/ComponentName.c > delete mode 100644 PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.c > delete mode 100644 > PcAtChipsetPkg/IsaAcpiDxe/PcatIsaAcpi.c > delete mode 100644 > PcAtChipsetPkg/8254TimerDxe/Timer.uni > delete mode 100644 > PcAtChipsetPkg/8254TimerDxe/TimerExtra.uni > delete mode 100644 > PcAtChipsetPkg/8259InterruptControllerDxe/Legacy8259.un > i > delete mode 100644 > PcAtChipsetPkg/8259InterruptControllerDxe/Legacy8259Ext > ra.uni > delete mode 100644 > PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.uni > delete mode 100644 > PcAtChipsetPkg/IsaAcpiDxe/IsaAcpiExtra.uni > > -- > 2.12.0.windows.1 > > > ^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: Linux Kernel soft hangs if 8259 PIC is not configured by BIOS: was (RE: [edk2-devel] [RFC][PATCH v1 0/1] PcAtChipsetPkg: Remove framework modules) 2019-05-28 16:36 Linux Kernel soft hangs if 8259 PIC is not configured by BIOS: was (RE: [edk2-devel] [RFC][PATCH v1 0/1] PcAtChipsetPkg: Remove framework modules) Kirkendall, Garrett 2019-05-28 17:06 ` Michael D Kinney @ 2019-05-28 17:17 ` Leif Lindholm 1 sibling, 0 replies; 4+ messages in thread From: Leif Lindholm @ 2019-05-28 17:17 UTC (permalink / raw) To: Kirkendall, Garrett Cc: devel@edk2.groups.io, hao.a.wu@intel.com, Ray Ni, Andrew Fish, Laszlo Ersek, Michael D Kinney Hi Garrett, Presumably you mean https://lkml.org/lkml/2019/5/9/304? I'm not familiar enough with the PC platform to stick my oar in, but Gleixner's counterquestion is an honest one - he does not see an obvious solution to the problem. And unless someone finds the problem interesting, this is very much a "patches welcome" sort of scenario. If Mike's suggestions don't help bring this forward, I would recommend you loop in all of the X86 maintainers, not just tglx. It may also be worth asking one of your Linux devs to have a look - Suravee Suthikulpanit is still at AMD, right? Best Regards, Leif On Tue, May 28, 2019 at 04:36:06PM +0000, Kirkendall, Garrett wrote: > Since this change will be removing the 8259 PIC driver, I'll mention > a quirk I ran into in the Linux Kernel up through at least 5.0 (and > I am sure later, I just didn't verify): > > Does anyone have pull in the Linux Kernel world to get them to take > this seriously? Hopefully someday soon end users will let legacy > devices, like the PIC, die a long overdue and graceful death. > > I sent the following to the responsible Linux Kernel developers and > quickly got back a "working as designed" response: > > ************************* > I am trying to boot a UEFI BIOS with minimal legacy hardware > support. The Linux kernel soft hangs when the PIC is not configured > by the BIOS because it is using IOAPIC. Hopefully, this provides > enough information. > > Observed under Ubuntu Server Linux 18.04 LTS with kernel 4.15.0, and with kernel compiled from source tag v5.0 > > Where it hangs: > Soft hang occurs in calibrate_APIC_clock(): https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/apic/apic.c#L805 > specific location of soft hang waiting for interrupts: https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/apic/apic.c#L854 > > > How it gets to the hang: > If 8259A PIC is not configured before kernel is launched, HPET IRQ 0 registration fails because probe_8259A returns PIC as not available and therefore interrupt descriptors 0-15 are not allocated. This happens when BIOS does not configure 8259A PIC because it uses IOAPIC. > > This sequence prevents allocating interrupts 0-15 unless PIC is configured before kernel starts. legacy_pic.init = init_8259A is not called before early_irq_init(): > early_irq_init(): https://github.com/torvalds/linux/blob/v5.0/init/main.c#L642 > initcnt = arch_probe_nr_irqs(): https://github.com/torvalds/linux/blob/v5.0/kernel/irq/irqdesc.c#L512 > return legacy_pic->probe(): https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/apic/vector.c#L656 > default_legacy_pic.probe: https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/i8259.c#L418 > probe_8259A(): https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/i8259.c#L301 > interrupt 0-15 descriptors not allocated and prevents IOAPIC interrupts 0-15: https://github.com/torvalds/linux/blob/v5.0/kernel/irq/irqdesc.c#L525 > > In this call is where init_8259A is called and PIC is initialized. However, interrupt descriptors for 0-15 were not allocated in early_irq_init() sequence, so descriptors are not available later. > init_IRQ() : https://github.com/torvalds/linux/blob/v5.0/init/main.c#L643 > > This sequence tries to register the HPET to irq0, but irq0 descriptor is not allocated by early_irq_init() sequence: > late_time_init(): https://github.com/torvalds/linux/blob/v5.0/init/main.c#L703 > late_time_init = x86_late_time_init: https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/time.c#L107 > x86_init.timers.timer_init(): https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/time.c#L92 > x86_init.timers.timer_init = hpet_time_init: https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/x86_init.c#L75 > setup_default_timer_irq(): https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/time.c#L83 > if (setup_irq(0, &irq0)): https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/time.c#L78 > > This gets called at some point after the above sequences, I couldn't track it all the way back to main.c easily: > x86_init.timers.setup_percpu_clockev = setup_boot_APIC_clock: https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/x86_init.c#L74 > https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/apic/apic.c#L961 > *** soft hang in calibrate_APIC_clock(): https://github.com/torvalds/linux/blob/v5.0/arch/x86/kernel/apic/apic.c#L854 > > ************************* > > GARRETT KIRKENDALL > SMTS Firmware Engineer | CTE > 7171 Southwest Parkway, Austin, TX 78735 USA > AMD facebook | amd.com > > -----Original Message----- > From: devel@edk2.groups.io <devel@edk2.groups.io> On Behalf Of Wu, Hao A via Groups.Io > Sent: Monday, May 27, 2019 1:37 AM > To: devel@edk2.groups.io > Cc: Hao A Wu <hao.a.wu@intel.com>; Ray Ni <ray.ni@intel.com>; Andrew Fish <afish@apple.com>; Laszlo Ersek <lersek@redhat.com>; Leif Lindholm <leif.lindholm@linaro.org>; Michael D Kinney <michael.d.kinney@intel.com> > Subject: [edk2-devel] [RFC][PATCH v1 0/1] PcAtChipsetPkg: Remove framework modules > > [CAUTION: External Email] > > ''' > Please note that this patch will be hold until all the below requirements are met: > > A. edk2-stable201905 is created; > B. OvmfPkg has drop its usage of the legacy ISA stack (which includes > the IsaAcpiDxe driver); > C. UefiPayloadPkg has been updated to drop its consume to the > 8259InterruptControllerDxe driver. > ''' > > This patch itself will not be sent to the mailing list and is only available at: > https://github.com/hwu25/edk2/tree/delete_framework > (https://github.com/hwu25/edk2/commit/84e58f89f6d03f9cc3399cced9d5d5529e06a416) > > > Below modules will be removed from PcAtChipsetPkg: > * PcAtChipsetPkg/8259InterruptControllerDxe/8259.inf > * PcAtChipsetPkg/8254TimerDxe/8254Timer.inf > * PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.inf > > They are considered legacy framework components and will no longer be used after the removal of IntelFramework[Module]Pkg. > > Also, the unused (after the modules being removed) PCDs will be deleted in package level DEC/UNI files. > > Cc: Ray Ni <ray.ni@intel.com> > Cc: Andrew Fish <afish@apple.com> > Cc: Laszlo Ersek <lersek@redhat.com> > Cc: Leif Lindholm <leif.lindholm@linaro.org> > Cc: Michael D Kinney <michael.d.kinney@intel.com> > > > Hao A Wu (1): > PcAtChipsetPkg: Remove framework modules > > PcAtChipsetPkg/PcAtChipsetPkg.dec | 58 -- > PcAtChipsetPkg/PcAtChipsetPkg.dsc | 5 +- > PcAtChipsetPkg/8254TimerDxe/8254Timer.inf | 42 -- > PcAtChipsetPkg/8259InterruptControllerDxe/8259.inf | 46 -- > PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.inf | 50 -- > PcAtChipsetPkg/8254TimerDxe/Timer.h | 185 ------ > PcAtChipsetPkg/8259InterruptControllerDxe/8259.h | 220 ------- > PcAtChipsetPkg/IsaAcpiDxe/PcatIsaAcpi.h | 269 --------- > PcAtChipsetPkg/8254TimerDxe/Timer.c | 401 ------------- > PcAtChipsetPkg/8259InterruptControllerDxe/8259.c | 622 -------------------- > PcAtChipsetPkg/IsaAcpiDxe/ComponentName.c | 301 ---------- > PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.c | 353 ----------- > PcAtChipsetPkg/IsaAcpiDxe/PcatIsaAcpi.c | 386 ------------ > PcAtChipsetPkg/8254TimerDxe/Timer.uni | 16 - > PcAtChipsetPkg/8254TimerDxe/TimerExtra.uni | 14 - > PcAtChipsetPkg/8259InterruptControllerDxe/Legacy8259.uni | 16 - > PcAtChipsetPkg/8259InterruptControllerDxe/Legacy8259Extra.uni | 14 - > PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.uni | 16 - > PcAtChipsetPkg/IsaAcpiDxe/IsaAcpiExtra.uni | 14 - > PcAtChipsetPkg/PcAtChipsetPkg.uni | 52 +- > 20 files changed, 2 insertions(+), 3078 deletions(-) delete mode 100644 PcAtChipsetPkg/8254TimerDxe/8254Timer.inf > delete mode 100644 PcAtChipsetPkg/8259InterruptControllerDxe/8259.inf > delete mode 100644 PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.inf > delete mode 100644 PcAtChipsetPkg/8254TimerDxe/Timer.h > delete mode 100644 PcAtChipsetPkg/8259InterruptControllerDxe/8259.h > delete mode 100644 PcAtChipsetPkg/IsaAcpiDxe/PcatIsaAcpi.h > delete mode 100644 PcAtChipsetPkg/8254TimerDxe/Timer.c > delete mode 100644 PcAtChipsetPkg/8259InterruptControllerDxe/8259.c > delete mode 100644 PcAtChipsetPkg/IsaAcpiDxe/ComponentName.c > delete mode 100644 PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.c > delete mode 100644 PcAtChipsetPkg/IsaAcpiDxe/PcatIsaAcpi.c > delete mode 100644 PcAtChipsetPkg/8254TimerDxe/Timer.uni > delete mode 100644 PcAtChipsetPkg/8254TimerDxe/TimerExtra.uni > delete mode 100644 PcAtChipsetPkg/8259InterruptControllerDxe/Legacy8259.uni > delete mode 100644 PcAtChipsetPkg/8259InterruptControllerDxe/Legacy8259Extra.uni > delete mode 100644 PcAtChipsetPkg/IsaAcpiDxe/IsaAcpi.uni > delete mode 100644 PcAtChipsetPkg/IsaAcpiDxe/IsaAcpiExtra.uni > > -- > 2.12.0.windows.1 > > > > ^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2019-05-28 17:31 UTC | newest] Thread overview: 4+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2019-05-28 16:36 Linux Kernel soft hangs if 8259 PIC is not configured by BIOS: was (RE: [edk2-devel] [RFC][PATCH v1 0/1] PcAtChipsetPkg: Remove framework modules) Kirkendall, Garrett 2019-05-28 17:06 ` Michael D Kinney 2019-05-28 17:31 ` Kirkendall, Garrett 2019-05-28 17:17 ` Leif Lindholm
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox