| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
When removing memory we need to remove the memory from the node
it was added to instead of looking up the node it should be in
in the device tree.
During testing we have seen scenarios where the affinity for a
LMB changes due to a partition migration or PRRN event. In these
cases the node the LMB exists in may not match the node the device
tree indicates it belongs in. This can lead to a system crash
when trying to DLPAR remove the LMB after a migration or PRRN
event. The current code looks up the node in the device tree to
remove the LMB from, the crash occurs when we try to offline this
node and it does not have any data, i.e. node_data[nid] == NULL.
36:mon> e
cpu 0x36: Vector: 300 (Data Access) at [c0000001828b7810]
pc: c00000000036d08c: try_offline_node+0x2c/0x1b0
lr: c0000000003a14ec: remove_memory+0xbc/0x110
sp: c0000001828b7a90
msr: 800000000280b033
dar: 9a28
dsisr: 40000000
current = 0xc0000006329c4c80
paca = 0xc000000007a55200 softe: 0 irq_happened: 0x01
pid = 76926, comm = kworker/u320:3
36:mon> t
[link register ] c0000000003a14ec remove_memory+0xbc/0x110
[c0000001828b7a90] c00000000006a1cc arch_remove_memory+0x9c/0xd0 (unreliable)
[c0000001828b7ad0] c0000000003a14e0 remove_memory+0xb0/0x110
[c0000001828b7b20] c0000000000c7db4 dlpar_remove_lmb+0x94/0x160
[c0000001828b7b60] c0000000000c8ef8 dlpar_memory+0x7e8/0xd10
[c0000001828b7bf0] c0000000000bf828 handle_dlpar_errorlog+0xf8/0x160
[c0000001828b7c60] c0000000000bf8cc pseries_hp_work_fn+0x3c/0xa0
[c0000001828b7c90] c000000000128cd8 process_one_work+0x298/0x5a0
[c0000001828b7d20] c000000000129068 worker_thread+0x88/0x620
[c0000001828b7dc0] c00000000013223c kthread+0x1ac/0x1c0
[c0000001828b7e30] c00000000000b45c ret_from_kernel_thread+0x5c/0x80
To resolve this we need to track the node a LMB belongs to when
it is added to the system so we can remove it from that node instead
of the node that the device tree indicates it should belong to.
Signed-off-by: Nathan Fontenot <[email protected]>
Signed-off-by: Michael Ellerman <[email protected]>
|
|
Replace the stack_trace_save*() functions with the new arch_stack_walk()
interfaces.
Signed-off-by: Thomas Gleixner <[email protected]>
Reviewed-by: Josh Poimboeuf <[email protected]>
Cc: Andy Lutomirski <[email protected]>
Cc: [email protected]
Cc: Steven Rostedt <[email protected]>
Cc: Alexander Potapenko <[email protected]>
Cc: Alexey Dobriyan <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Christoph Lameter <[email protected]>
Cc: Pekka Enberg <[email protected]>
Cc: [email protected]
Cc: David Rientjes <[email protected]>
Cc: Catalin Marinas <[email protected]>
Cc: Dmitry Vyukov <[email protected]>
Cc: Andrey Ryabinin <[email protected]>
Cc: [email protected]
Cc: Mike Rapoport <[email protected]>
Cc: Akinobu Mita <[email protected]>
Cc: Christoph Hellwig <[email protected]>
Cc: [email protected]
Cc: Robin Murphy <[email protected]>
Cc: Marek Szyprowski <[email protected]>
Cc: Johannes Thumshirn <[email protected]>
Cc: David Sterba <[email protected]>
Cc: Chris Mason <[email protected]>
Cc: Josef Bacik <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: Mike Snitzer <[email protected]>
Cc: Alasdair Kergon <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: [email protected]
Cc: Joonas Lahtinen <[email protected]>
Cc: Maarten Lankhorst <[email protected]>
Cc: [email protected]
Cc: David Airlie <[email protected]>
Cc: Jani Nikula <[email protected]>
Cc: Rodrigo Vivi <[email protected]>
Cc: Tom Zanussi <[email protected]>
Cc: Miroslav Benes <[email protected]>
Link: https://lkml.kernel.org/r/[email protected]
|
|
The following warning occurred on s390:
WARNING: CPU: 0 PID: 804 at kernel/locking/lockdep.c:1025 lockdep_register_key+0x30/0x150
This is because the check in static_obj() assumes that all memory within
[_stext, _end] belongs to static objects, which at least for s390 isn't
true. The init section is also part of this range, and freeing it allows
the buddy allocator to allocate memory from it. We have virt == phys for
the kernel on s390, so that such allocations would then have addresses
within the range [_stext, _end].
To fix this, introduce arch_is_kernel_initmem_freed(), similar to
arch_is_kernel_text/data(), and add it to the checks in static_obj().
This will always return 0 on architectures that do not define
arch_is_kernel_initmem_freed. On s390, it will return 1 if initmem has
been freed and the address is in the range [__init_begin, __init_end].
Signed-off-by: Gerald Schaefer <[email protected]>
Reviewed-by: Heiko Carstens <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
This patch adds support for relocating the kernel to a random address.
The random kernel offset is obtained from cpacf, using either TRNG, PRNO,
or KMC_PRNG, depending on supported MSA level.
KERNELOFFSET is added to vmcoreinfo, for crash --kaslr support.
Signed-off-by: Gerald Schaefer <[email protected]>
Reviewed-by: Philipp Rudo <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
With a relocatable kernel that could reside at any place in memory, code
and data that has to stay below 2 GB needs special handling.
This patch introduces .dma sections for such text, data and ex_table.
The sections will be part of the decompressor kernel, so they will not
be relocated and stay below 2 GB. Their location is passed over to the
decompressed / relocated kernel via the .boot.preserved.data section.
The duald and aste for control register setup also need to stay below
2 GB, so move the setup code from arch/s390/kernel/head64.S to
arch/s390/boot/head.S. The duct and linkage_stack could reside above
2 GB, but their content has to be preserved for the decompresed kernel,
so they are also moved into the .dma section.
The start and end address of the .dma sections is added to vmcoreinfo,
for crash support, to help debugging in case the kernel crashed there.
Signed-off-by: Gerald Schaefer <[email protected]>
Reviewed-by: Philipp Rudo <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
The sccbs for init/read/sdias/early have to be located below 2 GB, and
they are currently defined as a static buffer.
With a relocatable kernel that could reside at any place in memory, this
will no longer guarantee the location below 2 GB, so use a dynamic
GFP_DMA allocation instead.
The sclp_early_sccb buffer needs special handling, as it can be used
very early, and by both the decompressor and also the decompressed
kernel. Therefore, a fixed 4 KB buffer is introduced at 0x11000, the
former PARMAREA_END. The new PARMAREA_END is now 0x12000, and it is
renamed to HEAD_END, as it is rather the end of head.S and not the end
of the parmarea.
Signed-off-by: Gerald Schaefer <[email protected]>
Reviewed-by: Philipp Rudo <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
With a relocatable kernel that could reside at any place in memory, the
current logic for allocating a kprobes insn_page does not work. The
GFP_DMA allocated buffer might be more than 2 GB away from the kernel.
Use a static buffer for the insn_page instead.
Signed-off-by: Gerald Schaefer <[email protected]>
Reviewed-by: Philipp Rudo <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
With a relocatable kernel that could reside at any place in memory, the
storage size for the SYSCALL and PGM_CHECK handlers needs to be
increased from .long to .quad.
Signed-off-by: Gerald Schaefer <[email protected]>
Reviewed-by: Philipp Rudo <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
This patch adds support for building a relocatable kernel with -fPIE.
The kernel will be relocated to 0 early in the boot process.
Signed-off-by: Gerald Schaefer <[email protected]>
Reviewed-by: Philipp Rudo <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
Allow for userspace to use PCI MIO instructions.
Signed-off-by: Sebastian Ott <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
Allow users to disable usage of MIO instructions by specifying pci=nomio
at the kernel command line.
Signed-off-by: Sebastian Ott <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
Provide support for PCI I/O instructions that work on mapped IO addresses.
Signed-off-by: Sebastian Ott <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
This is a preparation patch for usage of new pci instructions.
No functional change.
Signed-off-by: Sebastian Ott <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
Provide a kernel parameter to force the usage of floating interrupts.
Signed-off-by: Sebastian Ott <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
Gather statistics to distinguish floating and directed interrupts.
Signed-off-by: Sebastian Ott <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
Improve /proc/interrupts on s390 to show statistics for individual
MSI interrupts.
Signed-off-by: Sebastian Ott <[email protected]>
Acked-by: Heiko Carstens <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
Up until now all interrupts on s390 have been floating. For MSI interrupts
we've used a global summary bit vector (with a bit for each function) and
a per-function interrupt bit vector (with a bit per MSI).
This patch introduces a new IRQ delivery mode: CPU directed interrupts.
In this new mode a per-CPU interrupt bit vector is used (with a bit per
MSI per function). Further it is now possible to direct an IRQ to a
specific CPU so we can finally support IRQ affinity.
If an interrupt can't be delivered because the appointed CPU is occupied
by a hypervisor the interrupt is delivered floating. For this a global
summary bit vector is used (with a bit per CPU).
Signed-off-by: Sebastian Ott <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
Provide the ability to create cachesize aligned interrupt vectors.
These will be used for per-CPU interrupt vectors.
Signed-off-by: Sebastian Ott <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
Rename and clarify the usage of the interrupt bit vectors. Also change
the array of the per-function bit vectors to be dynamically allocated.
Signed-off-by: Sebastian Ott <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
Add an extra parameter for airq handlers to recognize
floating vs. directed interrupts.
Signed-off-by: Sebastian Ott <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
Detect the adapter CPU directed interruption facility.
Signed-off-by: Sebastian Ott <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
Move everything interrupt related from pci.c to pci_irq.c.
No functional change.
Signed-off-by: Sebastian Ott <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
No users of pr_debug in that file.
Signed-off-by: Sebastian Ott <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
No point to keep that around.
Signed-off-by: Sebastian Ott <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
Provide an interface for userspace so it can find out if a machine is
capeable of doing secure boot. The interface is, for example, needed for
zipl so it can find out which file format it can/should write to disk.
Signed-off-by: Philipp Rudo <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
A kernel loaded via kexec_load cannot be verified. Thus disable kexec_load
systemcall in kernels which where IPLed securely. Use the IMA mechanism to
do so.
Signed-off-by: Philipp Rudo <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
Signed-off-by: Philipp Rudo <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
Add kernel signature verification to kexec_file. The verification is based
on module signature verification and works with kernel images signed via
scripts/sign-file.
Signed-off-by: Philipp Rudo <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
The leading 64 kB of a kernel image doesn't contain any data needed to boot
the new kernel when it was loaded via kexec_file. Thus kexec_file currently
strips them off before loading the image. Keep the leading 64 kB in order
to be able to pass a ipl_report to the next kernel.
Signed-off-by: Philipp Rudo <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
s390_image_load and s390_elf_load have the same code to load the different
components. Combine this functionality in one shared function.
While at it move kexec_file_update_kernel into the new function as well.
Signed-off-by: Philipp Rudo <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
Access the parmarea in head.S via a struct instead of individual offsets.
While at it make the fields in the parmarea .quads.
Signed-off-by: Philipp Rudo <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
Omit use of script/bin2c hack. Directly include into assembler file instead.
Signed-off-by: Philipp Rudo <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
The purgatory is compiled into the vmlinux and keept in memory all the time
during runtime. Thus any section not needed to load the purgatory
unnecessarily bloats up its foot print in file- and memorysize. Reduce the
purgatory size by stripping the unneeded sections from the purgatory.
This reduces the purgatories size by ~33%.
Signed-off-by: Philipp Rudo <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
To register data for the next kernel (command line, oldmem_base, etc.) the
current kernel needs to find the ELF segment that contains head.S. This is
currently done by checking ifor 'phdr->p_paddr == 0'. This works fine for
the current kernel build but in theory the first few pages could be
skipped. Make the detection more robust by checking if the entry point lies
within the segment.
Signed-off-by: Philipp Rudo <[email protected]>
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
When loading an ELF image via kexec_file the segment alignment is ignored
in the calculation for the load address of the next segment. When there are
multiple segments this can lead to segment overlap and thus load failure.
Signed-off-by: Philipp Rudo <[email protected]>
Fixes: 8be018827154 ("s390/kexec_file: Add ELF loader")
Signed-off-by: Martin Schwidefsky <[email protected]>
|
|
Let's use the correct validity number.
Fixes: 56019f9aca22 ("KVM: s390: vsie: Allow CRYCB FORMAT-2")
Signed-off-by: Pierre Morel <[email protected]>
Reviewed-by: Christian Borntraeger <[email protected]>
Message-Id: <[email protected]>
Acked-by: David Hildenbrand <[email protected]>
Signed-off-by: Christian Borntraeger <[email protected]>
|
|
When the guest do not have AP instructions nor Key management
we should return without shadowing the CRYCB.
We did not check correctly in the past.
Fixes: b10bd9a256ae ("s390: vsie: Use effective CRYCBD.31 to check CRYCBD validity")
Fixes: 6ee74098201b ("KVM: s390: vsie: allow CRYCB FORMAT-0")
Signed-off-by: Pierre Morel <[email protected]>
Reported-by: Christian Borntraeger <[email protected]>
Message-Id: <[email protected]>
Acked-by: David Hildenbrand <[email protected]>
Tested-by: Christian Borntraeger <[email protected]>
Reviewed-by: Christian Borntraeger <[email protected]>
Signed-off-by: Christian Borntraeger <[email protected]>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-nomadik into arm/soc
This modernizes the IXP4xx platform and adds initial Device Tree
Support. We migrate to MULTI_IRQ_HANDLER, bumps the IRQs to
offset 16, converts to SPARSE_IRQ, then we add proper subsystem
drivers in each subsystem for irqchip, GPIO and clocksource and
switch over to using these new drivers.
Next we modernize the NPE and QMGR drivers and push them down
into drivers/soc.
This has been tested on the IXP4xx NSLU2 and the Gateworks
GW2358-4.
* tag 'ixp4xx-for-armsoc' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-nomadik: (31 commits)
ARM: dts: Add queue manager and NPE to the IXP4xx DTSI
soc: ixp4xx: qmgr: Add DT probe code
soc: ixp4xx: qmgr: Add DT bindings for IXP4xx qmgr
soc: ixp4xx: npe: Add DT probe code
soc: ixp4xx: Add DT bindings for IXP4xx NPE
soc: ixp4xx: qmgr: Pass resources
soc: ixp4xx: Remove unused functions
soc: ixp4xx: Uninline several functions
soc: ixp4xx: npe: Pass addresses as resources
ARM: ixp4xx: Turn the QMGR into a platform device
ARM: ixp4xx: Turn the NPE into a platform device
ARM: ixp4xx: Move IXP4xx QMGR and NPE headers
ARM: ixp4xx: Move NPE and QMGR to drivers/soc
ARM: dts: Add some initial IXP4xx device trees
ARM: ixp4xx: Add device tree boot support
ARM: ixp4xx: Add DT bindings
gpio: ixp4xx: Add OF probing support
gpio: ixp4xx: Add DT bindings
clocksource/drivers/ixp4xx: Add OF initialization support
clocksource/drivers/ixp4xx: Add DT bindings
...
Signed-off-by: Olof Johansson <[email protected]>
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/horms/renesas into arm/soc
Renesas ARM Based SoC Updates for v5.2
* Power Management
- Fix a leaked reference by adding missing of_node_put
* R-Car M2-W (R8A77910) based Porter board
- Enable regulator quirk
* tag 'renesas-arm-soc-for-v5.2' of https://git.kernel.org/pub/scm/linux/kernel/git/horms/renesas:
ARM: shmobile: fix a leaked reference by adding missing of_node_put
ARM: shmobile: porter: enable R-Car Gen2 regulator quirk
Signed-off-by: Olof Johansson <[email protected]>
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/horms/renesas into arm/defconfig
Renesas ARM Based SoC Defconfig Updates for v5.2
shmobile and multi_v7 defconfigs
* Enable support for CFI NOR FLASH
shmobile defconfig
* Enable USB [EO]HCI HCD PLATFORM and PHY_RCAR_GEN3_USB2
* Refresh for v5.1-rc1
* tag 'renesas-arm-defconfig-for-v5.2' of https://git.kernel.org/pub/scm/linux/kernel/git/horms/renesas:
ARM: shmobile: Enable USB [EO]HCI HCD PLATFORM support in shmobile_defconfig
ARM: shmobile: Enable PHY_RCAR_GEN3_USB2 in shmobile_defconfig
ARM: multi_v7_defconfig: Enable support for CFI NOR FLASH
ARM: shmobile: defconfig: Enable support for CFI NOR FLASH
ARM: shmobile: defconfig: Refresh for v5.1-rc1
Signed-off-by: Olof Johansson <[email protected]>
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/horms/renesas into arm/defconfig
Renesas ARM64 Based SoC Defconfig Updates for v5.2
+ Enable support for RX-8571/RX-8581 RTC
* tag 'renesas-arm64-defconfig-for-v5.2' of https://git.kernel.org/pub/scm/linux/kernel/git/horms/renesas:
arm64: defconfig: enable RX-8581 config option
Signed-off-by: Olof Johansson <[email protected]>
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/sunxi/linux into arm/defconfig
Allwinner arm64 defconfig changes for 5.2
Just a single patch to enable our SPI controller on the arm64 defconfig.
* tag 'sunxi-config64-for-5.2' of https://git.kernel.org/pub/scm/linux/kernel/git/sunxi/linux:
arm64: defconfig: Enable SPI_SUN6I
Signed-off-by: Olof Johansson <[email protected]>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux into arm/defconfig
arm64: tegra: Default configuration updates for v5.2-rc1
These patches enable PWM fan and Tegra HDA support in the 64-bit ARM
default configuration, so that these features are enabled by default.
* tag 'tegra-for-5.2-arm64-defconfig' of git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux:
arm64: defconfig: Add PWM Fan support
arm64: defconfig: Enable Tegra HDA support
Signed-off-by: Olof Johansson <[email protected]>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux into arm/defconfig
ARM: tegra: Default configuration updates for v5.2-rc1
Enable Trusted Foundations support in the default configurations for
Tegra and multi-v7. This is necessary because the symbol is no longer
selected by default.
* tag 'tegra-for-5.2-arm-defconfig' of git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux:
ARM: Enable Trusted Foundations for multiplatform ARM v7
ARM: tegra: Enable Trusted Foundations by default
ARM: tegra: Update default configuration for v5.1-rc1
Signed-off-by: Olof Johansson <[email protected]>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux into arm/drivers
ARM: tegra: Core changes for v5.2-rc1
One of these patches enables PM by default on 32-bit ARM, following the
same default that we already have on 64-bit ARM. The other patch fixes a
cosmetic issue in the cpuidle driver for Tegra20 and Tegra30.
* tag 'tegra-for-5.2-arm-soc' of git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux:
ARM: tegra: cpuidle: Handle tick broadcasting within cpuidle core on Tegra20/30
ARM: tegra: enforce PM requirement
Signed-off-by: Olof Johansson <[email protected]>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux into arm/drivers
memory: tegra: Changes for v5.2-rc1
These are a set of fixes for various issues related to the Tegra memory
controller.
* tag 'tegra-for-5.2-memory' of git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux:
memory: tegra: Fix a typos for "fdcdwr2" mc client
Revert "ARM: tegra: Restore memory arbitration on resume from LP1 on Tegra30+"
memory: tegra: Replace readl-writel with mc_readl-mc_writel
memory: tegra: Fix integer overflow on tick value calculation
memory: tegra: Fix missed registers values latching
memory: tegra: Properly spell "tegra"
memory: tegra: Make terga20_mc_reset_ops static
Signed-off-by: Olof Johansson <[email protected]>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux into arm/soc
firmware: tegra: Changes for v5.2-rc1
This set of changes includes improvements for Trusted Foundations and
also moves the source files for this support into the standard location
under drivers/firmware.
* tag 'tegra-for-5.2-firmware' of git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux:
firmware: Move Trusted Foundations support
ARM: tegra: Sort dependencies alphabetically
ARM: tegra: Add firmware calls required for suspend-resume on Tegra30
ARM: tegra: Always boot CPU in ARM-mode
ARM: tegra: Don't apply CPU erratas in insecure mode
ARM: tegra: Set up L2 cache using Trusted Foundations firmware
ARM: trusted_foundations: Provide information about whether firmware is registered
ARM: trusted_foundations: Make prepare_idle call to take mode argument
ARM: trusted_foundations: Support L2 cache maintenance
Signed-off-by: Olof Johansson <[email protected]>
|
|
Currently perf callchain doesn't work well with ORC unwinder
when sampling from trace point. We'll get useless in kernel callchain
like this:
perf 6429 [000] 22.498450: kmem:mm_page_alloc: page=0x176a17 pfn=1534487 order=0 migratetype=0 gfp_flags=GFP_KERNEL
ffffffffbe23e32e __alloc_pages_nodemask+0x22e (/lib/modules/5.1.0-rc3+/build/vmlinux)
7efdf7f7d3e8 __poll+0x18 (/usr/lib64/libc-2.28.so)
5651468729c1 [unknown] (/usr/bin/perf)
5651467ee82a main+0x69a (/usr/bin/perf)
7efdf7eaf413 __libc_start_main+0xf3 (/usr/lib64/libc-2.28.so)
5541f689495641d7 [unknown] ([unknown])
The root cause is that, for trace point events, it doesn't provide a
real snapshot of the hardware registers. Instead perf tries to get
required caller's registers and compose a fake register snapshot
which suppose to contain enough information for start a unwinding.
However without CONFIG_FRAME_POINTER, if failed to get caller's BP as the
frame pointer, so current frame pointer is returned instead. We get
a invalid register combination which confuse the unwinder, and end the
stacktrace early.
So in such case just don't try dump BP, and let the unwinder start
directly when the register is not a real snapshot. Use SP
as the skip mark, unwinder will skip all the frames until it meet
the frame of the trace point caller.
Tested with frame pointer unwinder and ORC unwinder, this makes perf
callchain get the full kernel space stacktrace again like this:
perf 6503 [000] 1567.570191: kmem:mm_page_alloc: page=0x16c904 pfn=1493252 order=0 migratetype=0 gfp_flags=GFP_KERNEL
ffffffffb523e2ae __alloc_pages_nodemask+0x22e (/lib/modules/5.1.0-rc3+/build/vmlinux)
ffffffffb52383bd __get_free_pages+0xd (/lib/modules/5.1.0-rc3+/build/vmlinux)
ffffffffb52fd28a __pollwait+0x8a (/lib/modules/5.1.0-rc3+/build/vmlinux)
ffffffffb521426f perf_poll+0x2f (/lib/modules/5.1.0-rc3+/build/vmlinux)
ffffffffb52fe3e2 do_sys_poll+0x252 (/lib/modules/5.1.0-rc3+/build/vmlinux)
ffffffffb52ff027 __x64_sys_poll+0x37 (/lib/modules/5.1.0-rc3+/build/vmlinux)
ffffffffb500418b do_syscall_64+0x5b (/lib/modules/5.1.0-rc3+/build/vmlinux)
ffffffffb5a0008c entry_SYSCALL_64_after_hwframe+0x44 (/lib/modules/5.1.0-rc3+/build/vmlinux)
7f71e92d03e8 __poll+0x18 (/usr/lib64/libc-2.28.so)
55a22960d9c1 [unknown] (/usr/bin/perf)
55a22958982a main+0x69a (/usr/bin/perf)
7f71e9202413 __libc_start_main+0xf3 (/usr/lib64/libc-2.28.so)
5541f689495641d7 [unknown] ([unknown])
Co-developed-by: Josh Poimboeuf <[email protected]>
Signed-off-by: Kairui Song <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Alexei Starovoitov <[email protected]>
Cc: Arnaldo Carvalho de Melo <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Dave Young <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
The digicolor platform has three UARTs, but the Kconfig.debug
file explicitly lists port zero as the one to be used for the
console, while not providing any default values.
This can get an automated randconfig build stuck in a loop
waiting for the user to input the number. As we already know
the physical address, this patch provides that number as
default, along with a reasonable default value for the virtual
address.
Signed-off-by: Arnd Bergmann <[email protected]>
Reviewed-by: Baruch Siach <[email protected]>
Signed-off-by: Olof Johansson <[email protected]>
|
|
Commit 7ee7ef24d02d ("scsi: arm64: defconfig: enable configs for Hisilicon ufs")
set 'CONFIG_SCSI_UFS_HISI=y', but the configs it depends
on
(CONFIG_SCSI_HFSHCD_PLATFORM && CONFIG_SCSI_UFSHCD)
were left to being built as modules.
Commit 1f4fa50dd48f ("arm64: defconfig: Regenerate for v4.20") "fixed"
that by reverting to 'CONFIG_SCSI_UFS_HISI=m'.
Thing is, if the rootfs is stored in the on-board flash (which
is the "canonical" way of doing things), we either need these drivers
to be built-in, or we need to fiddle with an initramfs to access that
flash and eventually load the modules installed over there.
The former is the easiest, do that.
Signed-off-by: Valentin Schneider <[email protected]>
Reviewed-by: Leo Yan <[email protected]>
Signed-off-by: Olof Johansson <[email protected]>
|