mbox series

[0/5] arch_topology: Build cacheinfo from primary CPU

Message ID 20221108110424.166896-1-pierre.gondois@arm.com (mailing list archive)
Headers show
Series arch_topology: Build cacheinfo from primary CPU | expand

Message

Pierre Gondois Nov. 8, 2022, 11:04 a.m. UTC
[1] and [2] build the CPU topology from the cacheinfo information for
both DT/ACPI based systems and remove (struct cpu_topology).llc_id
which was used by ACPI only.

Creating the cacheinfo for secondary CPUs is done during early boot.
Preemption and interrupts are disabled at this stage. On PREEMPT_RT
kernels, allocating memory (and parsing the PPTT table for ACPI based
systems) triggers a:
  'BUG: sleeping function called from invalid context' [4]

To prevent this bug, allocate the cacheinfo from the primary CPU when
preemption and interrupts are enabled and before booting secondary
CPUs. The cache levels/leaves are computed from DT/ACPI PPTT information
only, without relying on the arm64 CLIDR_EL1 register.
If no cache information is found in the DT/ACPI PPTT, then fallback
to the current state, triggering [4] on PREEMPT_RT kernels.

Patches to update the arm64 device trees that have incomplete cacheinfo
(mostly for missing the 'cache-level' or 'cache-unified' property)
have been sent at [3].

Tested platforms:
- ACPI + PPTT: Ampere Altra, Ampere eMAG, Cavium ThunderX2,
  Kunpeng 920, Juno-r2
- DT: rb5, db845c, Juno-r2

[1] https://lore.kernel.org/all/20220704101605.1318280-1-sudeep.holla@arm.com/
[2] https://lore.kernel.org/all/20220720-arch_topo_fixes-v3-0-43d696288e84@arm.com/
[3] https://lore.kernel.org/all/20221107155825.1644604-1-pierre.gondois@arm.com/
[4] On an Ampere Altra, with PREEMPT_RT kernel based on v6.0.0-rc4:


[    7.560791] BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:46
[    7.560794] in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 0, name: swapper/111
[    7.560796] preempt_count: 1, expected: 0
[    7.560797] RCU nest depth: 1, expected: 1
[    7.560799] 3 locks held by swapper/111/0:
[    7.560800]  #0: ffff403e406cae98 (&pcp->lock){+.+.}-{3:3}, at: get_page_from_freelist+0x218/0x12c8
[    7.560811]  #1: ffffc5f8ed09f8e8 (rcu_read_lock){....}-{1:3}, at: rt_spin_trylock+0x48/0xf0
[    7.560820]  #2: ffff403f400b4fd8 (&zone->lock){+.+.}-{3:3}, at: rmqueue_bulk+0x64/0xa80
[    7.560824] irq event stamp: 0
[    7.560825] hardirqs last  enabled at (0): [<0000000000000000>] 0x0
[    7.560827] hardirqs last disabled at (0): [<ffffc5f8e9f7d594>] copy_process+0x5dc/0x1ab8
[    7.560830] softirqs last  enabled at (0): [<ffffc5f8e9f7d594>] copy_process+0x5dc/0x1ab8
[    7.560833] softirqs last disabled at (0): [<0000000000000000>] 0x0
[    7.560834] Preemption disabled at:
[    7.560835] [<ffffc5f8e9fd3c28>] migrate_enable+0x30/0x130
[    7.560838] CPU: 111 PID: 0 Comm: swapper/111 Tainted: G        W          6.0.0-rc4-[...]
[    7.560841] Call trace:
[...]
[    7.560870]  __kmalloc+0xbc/0x1e8
[    7.560873]  detect_cache_attributes+0x2d4/0x5f0
[    7.560876]  update_siblings_masks+0x30/0x368
[    7.560880]  store_cpu_topology+0x78/0xb8
[    7.560883]  secondary_start_kernel+0xd0/0x198
[    7.560885]  __secondary_switched+0xb0/0xb4

Pierre Gondois (5):
  cacheinfo: Use riscv's init_cache_level() as generic OF implem
  cacheinfo: Return error code in init_of_cache_level()
  ACPI: PPTT: Remove acpi_find_cache_levels()
  ACPI: PPTT: Update acpi_find_last_cache_level() to
    acpi_get_cache_info()
  arch_topology: Build cacheinfo from primary CPU

 arch/arm64/kernel/cacheinfo.c |   9 ++-
 arch/riscv/kernel/cacheinfo.c |  39 +------------
 drivers/acpi/pptt.c           |  86 +++++++++++++++++------------
 drivers/base/arch_topology.c  |  10 +++-
 drivers/base/cacheinfo.c      | 101 ++++++++++++++++++++++++++++++----
 include/linux/cacheinfo.h     |  10 +++-
 6 files changed, 164 insertions(+), 91 deletions(-)

Comments

Pierre Gondois Nov. 8, 2022, 11:11 a.m. UTC | #1
+ Rob Herring
+ Jeremy Linton

On 11/8/22 12:04, Pierre Gondois wrote:
> [1] and [2] build the CPU topology from the cacheinfo information for
> both DT/ACPI based systems and remove (struct cpu_topology).llc_id
> which was used by ACPI only.
> 
> Creating the cacheinfo for secondary CPUs is done during early boot.
> Preemption and interrupts are disabled at this stage. On PREEMPT_RT
> kernels, allocating memory (and parsing the PPTT table for ACPI based
> systems) triggers a:
>    'BUG: sleeping function called from invalid context' [4]
> 
> To prevent this bug, allocate the cacheinfo from the primary CPU when
> preemption and interrupts are enabled and before booting secondary
> CPUs. The cache levels/leaves are computed from DT/ACPI PPTT information
> only, without relying on the arm64 CLIDR_EL1 register.
> If no cache information is found in the DT/ACPI PPTT, then fallback
> to the current state, triggering [4] on PREEMPT_RT kernels.
> 
> Patches to update the arm64 device trees that have incomplete cacheinfo
> (mostly for missing the 'cache-level' or 'cache-unified' property)
> have been sent at [3].
> 
> Tested platforms:
> - ACPI + PPTT: Ampere Altra, Ampere eMAG, Cavium ThunderX2,
>    Kunpeng 920, Juno-r2
> - DT: rb5, db845c, Juno-r2
> 
> [1] https://lore.kernel.org/all/20220704101605.1318280-1-sudeep.holla@arm.com/
> [2] https://lore.kernel.org/all/20220720-arch_topo_fixes-v3-0-43d696288e84@arm.com/
> [3] https://lore.kernel.org/all/20221107155825.1644604-1-pierre.gondois@arm.com/
> [4] On an Ampere Altra, with PREEMPT_RT kernel based on v6.0.0-rc4:
> 
> 
> [    7.560791] BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:46
> [    7.560794] in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 0, name: swapper/111
> [    7.560796] preempt_count: 1, expected: 0
> [    7.560797] RCU nest depth: 1, expected: 1
> [    7.560799] 3 locks held by swapper/111/0:
> [    7.560800]  #0: ffff403e406cae98 (&pcp->lock){+.+.}-{3:3}, at: get_page_from_freelist+0x218/0x12c8
> [    7.560811]  #1: ffffc5f8ed09f8e8 (rcu_read_lock){....}-{1:3}, at: rt_spin_trylock+0x48/0xf0
> [    7.560820]  #2: ffff403f400b4fd8 (&zone->lock){+.+.}-{3:3}, at: rmqueue_bulk+0x64/0xa80
> [    7.560824] irq event stamp: 0
> [    7.560825] hardirqs last  enabled at (0): [<0000000000000000>] 0x0
> [    7.560827] hardirqs last disabled at (0): [<ffffc5f8e9f7d594>] copy_process+0x5dc/0x1ab8
> [    7.560830] softirqs last  enabled at (0): [<ffffc5f8e9f7d594>] copy_process+0x5dc/0x1ab8
> [    7.560833] softirqs last disabled at (0): [<0000000000000000>] 0x0
> [    7.560834] Preemption disabled at:
> [    7.560835] [<ffffc5f8e9fd3c28>] migrate_enable+0x30/0x130
> [    7.560838] CPU: 111 PID: 0 Comm: swapper/111 Tainted: G        W          6.0.0-rc4-[...]
> [    7.560841] Call trace:
> [...]
> [    7.560870]  __kmalloc+0xbc/0x1e8
> [    7.560873]  detect_cache_attributes+0x2d4/0x5f0
> [    7.560876]  update_siblings_masks+0x30/0x368
> [    7.560880]  store_cpu_topology+0x78/0xb8
> [    7.560883]  secondary_start_kernel+0xd0/0x198
> [    7.560885]  __secondary_switched+0xb0/0xb4
> 
> Pierre Gondois (5):
>    cacheinfo: Use riscv's init_cache_level() as generic OF implem
>    cacheinfo: Return error code in init_of_cache_level()
>    ACPI: PPTT: Remove acpi_find_cache_levels()
>    ACPI: PPTT: Update acpi_find_last_cache_level() to
>      acpi_get_cache_info()
>    arch_topology: Build cacheinfo from primary CPU
> 
>   arch/arm64/kernel/cacheinfo.c |   9 ++-
>   arch/riscv/kernel/cacheinfo.c |  39 +------------
>   drivers/acpi/pptt.c           |  86 +++++++++++++++++------------
>   drivers/base/arch_topology.c  |  10 +++-
>   drivers/base/cacheinfo.c      | 101 ++++++++++++++++++++++++++++++----
>   include/linux/cacheinfo.h     |  10 +++-
>   6 files changed, 164 insertions(+), 91 deletions(-)
>
Conor Dooley Nov. 8, 2022, 2:04 p.m. UTC | #2
On Tue, Nov 08, 2022 at 12:04:16PM +0100, Pierre Gondois wrote:
> [1] and [2] build the CPU topology from the cacheinfo information for
> both DT/ACPI based systems and remove (struct cpu_topology).llc_id
> which was used by ACPI only.
> 
> Creating the cacheinfo for secondary CPUs is done during early boot.
> Preemption and interrupts are disabled at this stage. On PREEMPT_RT
> kernels, allocating memory (and parsing the PPTT table for ACPI based
> systems) triggers a:
>   'BUG: sleeping function called from invalid context' [4]
> 
> To prevent this bug, allocate the cacheinfo from the primary CPU when
> preemption and interrupts are enabled and before booting secondary
> CPUs. The cache levels/leaves are computed from DT/ACPI PPTT information
> only, without relying on the arm64 CLIDR_EL1 register.
> If no cache information is found in the DT/ACPI PPTT, then fallback
> to the current state, triggering [4] on PREEMPT_RT kernels.
> 
> Patches to update the arm64 device trees that have incomplete cacheinfo
> (mostly for missing the 'cache-level' or 'cache-unified' property)
> have been sent at [3].

FWIW checkpatch whinges a bit at your final two patches [a]. The RISC-V
code movement looks okay though, no new warnings and the like.

[a] - https://patchwork.kernel.org/project/linux-riscv/list/?series=693175

> 
> Tested platforms:
> - ACPI + PPTT: Ampere Altra, Ampere eMAG, Cavium ThunderX2,
>   Kunpeng 920, Juno-r2
> - DT: rb5, db845c, Juno-r2
> 
> [1] https://lore.kernel.org/all/20220704101605.1318280-1-sudeep.holla@arm.com/
> [2] https://lore.kernel.org/all/20220720-arch_topo_fixes-v3-0-43d696288e84@arm.com/
> [3] https://lore.kernel.org/all/20221107155825.1644604-1-pierre.gondois@arm.com/
> [4] On an Ampere Altra, with PREEMPT_RT kernel based on v6.0.0-rc4:
> 
> 
> [    7.560791] BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:46
> [    7.560794] in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 0, name: swapper/111
> [    7.560796] preempt_count: 1, expected: 0
> [    7.560797] RCU nest depth: 1, expected: 1
> [    7.560799] 3 locks held by swapper/111/0:
> [    7.560800]  #0: ffff403e406cae98 (&pcp->lock){+.+.}-{3:3}, at: get_page_from_freelist+0x218/0x12c8
> [    7.560811]  #1: ffffc5f8ed09f8e8 (rcu_read_lock){....}-{1:3}, at: rt_spin_trylock+0x48/0xf0
> [    7.560820]  #2: ffff403f400b4fd8 (&zone->lock){+.+.}-{3:3}, at: rmqueue_bulk+0x64/0xa80
> [    7.560824] irq event stamp: 0
> [    7.560825] hardirqs last  enabled at (0): [<0000000000000000>] 0x0
> [    7.560827] hardirqs last disabled at (0): [<ffffc5f8e9f7d594>] copy_process+0x5dc/0x1ab8
> [    7.560830] softirqs last  enabled at (0): [<ffffc5f8e9f7d594>] copy_process+0x5dc/0x1ab8
> [    7.560833] softirqs last disabled at (0): [<0000000000000000>] 0x0
> [    7.560834] Preemption disabled at:
> [    7.560835] [<ffffc5f8e9fd3c28>] migrate_enable+0x30/0x130
> [    7.560838] CPU: 111 PID: 0 Comm: swapper/111 Tainted: G        W          6.0.0-rc4-[...]
> [    7.560841] Call trace:
> [...]
> [    7.560870]  __kmalloc+0xbc/0x1e8
> [    7.560873]  detect_cache_attributes+0x2d4/0x5f0
> [    7.560876]  update_siblings_masks+0x30/0x368
> [    7.560880]  store_cpu_topology+0x78/0xb8
> [    7.560883]  secondary_start_kernel+0xd0/0x198
> [    7.560885]  __secondary_switched+0xb0/0xb4
> 
> Pierre Gondois (5):
>   cacheinfo: Use riscv's init_cache_level() as generic OF implem
>   cacheinfo: Return error code in init_of_cache_level()
>   ACPI: PPTT: Remove acpi_find_cache_levels()
>   ACPI: PPTT: Update acpi_find_last_cache_level() to
>     acpi_get_cache_info()
>   arch_topology: Build cacheinfo from primary CPU
> 
>  arch/arm64/kernel/cacheinfo.c |   9 ++-
>  arch/riscv/kernel/cacheinfo.c |  39 +------------
>  drivers/acpi/pptt.c           |  86 +++++++++++++++++------------
>  drivers/base/arch_topology.c  |  10 +++-
>  drivers/base/cacheinfo.c      | 101 ++++++++++++++++++++++++++++++----
>  include/linux/cacheinfo.h     |  10 +++-
>  6 files changed, 164 insertions(+), 91 deletions(-)
> 
> -- 
> 2.25.1
> 
> 
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel