Re: [PATCH v1 1/3] LoongArch: Implement CONFIG_THREAD_INFO_IN_TASK

[Huacai Chen]

Hi, Tiezhu,

There are still some issues.

On Mon, Jun 8, 2026 at 6:35 PM Tiezhu Yang <yangtiezhu@xxxxxxxxxxx> wrote:
>
> Like other architectures such as x86, arm64, riscv, powerpc and s390,
> select THREAD_INFO_IN_TASK for LoongArch to move thread_info off the
> stack into task_struct. This follows modern kernel standards and also
> makes the system more secure.
>
> With this patch, thread_info is included in task_struct at an offset
> of 0 instead of being placed at the bottom of the kernel stack. Thus,
> the $tp register points to both thread_info and task_struct.
>
> To support this, introduce a per-CPU variable cpu_tasks to store the
> pointer to the current task_struct. This decouples the recovery of
> the $tp register from the stack pointer during exception entry.
>
> Then initialize cpu_tasks for the primary and secondary CPUs during
> arch-specific setup and SMP boot paths. To eliminate the dangerous
> windows during the early initialization where the cpu_tasks remains
> uninitialized, set_current() is invoked as early as possible in both
> setup_arch() (right after unwind_init) and start_secondary() (right
> after cpu_probe). This ensures the $tp recovery barrier is armed in
> case any early boot exceptions or kernel panics occur.
>
> Modify SAVE_SOME and handle_syscall to restore the $tp register from
> cpu_tasks, and also use the la_abs absolute addressing for cpu_tasks
> access in assembly to bypass the relocation limits within exception
> handling sections. By advancing the preservation of u0 in SAVE_SOME,
> reuse the PERCPU_BASE_KS value in u0 for the cpu_tasks calculation,
> effectively eliminating a duplicate csrrd instruction execution on
> SMP platforms.
>
> Remove the obsolete next_ti argument from __switch_to(), which shifts
> the remaining arguments ahead in the calling convention (sched_ra from
> a3 to a2, and sched_cfa from a4 to a3). Also, update __switch_to() to
> fetch the kernel stack pointer using (TASK_STACK - TASK_STRUCT_OFFSET)
> instead of a direct offset, which correctly neutralizes the structural
> pointer bias on 32-bit platforms while cleanly falling back to 0 on
> 64-bit systems. This ensures full multi-architecture compatibility
> while __switch_to() directly updates $tp from a1 for efficiency.
>
> Signed-off-by: Tiezhu Yang <yangtiezhu@xxxxxxxxxxx>
> ---
>  .../core/thread-info-in-task/arch-support.txt |  2 +-
>  arch/loongarch/Kconfig                        |  1 +
>  arch/loongarch/include/asm/current.h          | 30 +++++++++++++++++++
>  arch/loongarch/include/asm/smp.h              |  3 +-
>  arch/loongarch/include/asm/stackframe.h       |  9 ++++--
>  arch/loongarch/include/asm/switch_to.h        |  6 ++--
>  arch/loongarch/include/asm/thread_info.h      | 10 -------
>  arch/loongarch/kernel/asm-offsets.c           |  6 ++--
>  arch/loongarch/kernel/entry.S                 |  7 +++--
>  arch/loongarch/kernel/head.S                  | 18 ++++++-----
>  arch/loongarch/kernel/process.c               |  3 ++
>  arch/loongarch/kernel/relocate.c              |  2 +-
>  arch/loongarch/kernel/setup.c                 |  1 +
>  arch/loongarch/kernel/smp.c                   |  6 ++--
>  arch/loongarch/kernel/switch.S                | 13 ++++----
>  15 files changed, 79 insertions(+), 38 deletions(-)
>  create mode 100644 arch/loongarch/include/asm/current.h
First of all, you should protect show_backtrace() in
arch/loongarch/kernel/traps.c with try_get_task_stack() and
put_task_stack(), which is mentioned in init/Kconfig. You can use
ARM64's dump_backtrace() as an example.


>
> diff --git a/Documentation/features/core/thread-info-in-task/arch-support.txt b/Documentation/features/core/thread-info-in-task/arch-support.txt
> index f3d744c76061..e26efdfbb6b4 100644
> --- a/Documentation/features/core/thread-info-in-task/arch-support.txt
> +++ b/Documentation/features/core/thread-info-in-task/arch-support.txt
> @@ -12,7 +12,7 @@
>      |       arm64: |  ok  |
>      |        csky: | TODO |
>      |     hexagon: | TODO |
> -    |   loongarch: | TODO |
> +    |   loongarch: |  ok  |
>      |        m68k: | TODO |
>      |  microblaze: | TODO |
>      |        mips: | TODO |
> diff --git a/arch/loongarch/Kconfig b/arch/loongarch/Kconfig
> index 606597da46b8..cf8d3cf91814 100644
> --- a/arch/loongarch/Kconfig
> +++ b/arch/loongarch/Kconfig
> @@ -210,6 +210,7 @@ config LOONGARCH
>         select SYSCTL_ARCH_UNALIGN_NO_WARN
>         select SYSCTL_EXCEPTION_TRACE
>         select SWIOTLB if 64BIT
> +       select THREAD_INFO_IN_TASK
>         select TRACE_IRQFLAGS_SUPPORT
>         select USE_PERCPU_NUMA_NODE_ID
>         select USER_STACKTRACE_SUPPORT
> diff --git a/arch/loongarch/include/asm/current.h b/arch/loongarch/include/asm/current.h
> new file mode 100644
> index 000000000000..1ee8517c0291
> --- /dev/null
> +++ b/arch/loongarch/include/asm/current.h
> @@ -0,0 +1,30 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +#ifndef __ASM_LOONGARCH_CURRENT_H
> +#define __ASM_LOONGARCH_CURRENT_H
> +
> +#include <linux/compiler.h>
> +#include <asm/percpu.h>
It is better to move this file inside #ifndef __ASSEMBLER__ as x86
does, this can avoid some potential build errors in case this file is
included in .S files.

> +
> +#ifndef __ASSEMBLER__
> +
> +struct task_struct;
> +
> +DECLARE_PER_CPU(struct task_struct *, cpu_tasks);
> +
> +register struct task_struct *current_thread_pointer __asm__("$tp");
> +
> +static __always_inline struct task_struct *get_current(void)
> +{
> +       return current_thread_pointer;
> +}
> +
> +#define current get_current()
> +
> +static __always_inline void set_current(struct task_struct *task)
> +{
> +       __this_cpu_write(cpu_tasks, task);
> +}
> +
> +#endif /* __ASSEMBLER__ */
> +
> +#endif /* __ASM_LOONGARCH_CURRENT_H */
> diff --git a/arch/loongarch/include/asm/smp.h b/arch/loongarch/include/asm/smp.h
> index 3a47f52959a8..761cc0016df8 100644
> --- a/arch/loongarch/include/asm/smp.h
> +++ b/arch/loongarch/include/asm/smp.h
> @@ -82,7 +82,8 @@ struct seq_file;
>
>  struct secondary_data {
>         unsigned long stack;
> -       unsigned long thread_info;
> +       unsigned long task;
> +       unsigned long offset;
>  };
>  extern struct secondary_data cpuboot_data;
>
> diff --git a/arch/loongarch/include/asm/stackframe.h b/arch/loongarch/include/asm/stackframe.h
> index ecc8e50fffa8..b031152e0fc8 100644
> --- a/arch/loongarch/include/asm/stackframe.h
> +++ b/arch/loongarch/include/asm/stackframe.h
> @@ -191,10 +191,15 @@
>         andi    t0, t0, 0x3     /* extract pplv bit */
>         beqz    t0, 9f
>
> -       LONG_LI tp, ~_THREAD_MASK
> -       and     tp, tp, sp
>         cfi_st  u0, PT_R21, \docfi
>         csrrd   u0, PERCPU_BASE_KS
> +
> +       la_abs      t1, cpu_tasks
> +#ifdef CONFIG_SMP
> +       LONG_ADD    t1, t1, u0
> +#endif
> +       LONG_L      tp, t1, 0
> +
>  9:
>  #ifdef CONFIG_KGDB
>         li.w    t0, CSR_CRMD_WE
> diff --git a/arch/loongarch/include/asm/switch_to.h b/arch/loongarch/include/asm/switch_to.h
> index 5b225aff3ba2..07a4c5909779 100644
> --- a/arch/loongarch/include/asm/switch_to.h
> +++ b/arch/loongarch/include/asm/switch_to.h
> @@ -15,7 +15,6 @@ struct task_struct;
>   * __switch_to - switch execution of a task
>   * @prev:      The task previously executed.
>   * @next:      The task to begin executing.
> - * @next_ti:   task_thread_info(next).
>   * @sched_ra:  __schedule return address.
>   * @sched_cfa: __schedule call frame address.
>   *
> @@ -23,7 +22,7 @@ struct task_struct;
>   * the context of next. Returns prev.
>   */
>  extern asmlinkage struct task_struct *__switch_to(struct task_struct *prev,
> -                       struct task_struct *next, struct thread_info *next_ti,
> +                       struct task_struct *next,
>                         void *sched_ra, void *sched_cfa);
>
>  /*
> @@ -37,7 +36,8 @@ do {                                                                          \
>         lose_fpu_inatomic(1, prev);                                             \
>         lose_lbt_inatomic(1, prev);                                             \
>         hw_breakpoint_thread_switch(next);                                      \
> -       (last) = __switch_to(prev, next, task_thread_info(next),                \
> +       set_current(next);                                                      \
> +       (last) = __switch_to(prev, next,                                        \
>                  __builtin_return_address(0), __builtin_frame_address(0));      \
>  } while (0)
>
> diff --git a/arch/loongarch/include/asm/thread_info.h b/arch/loongarch/include/asm/thread_info.h
> index 4d7117fcdc78..41eabe4fb647 100644
> --- a/arch/loongarch/include/asm/thread_info.h
> +++ b/arch/loongarch/include/asm/thread_info.h
> @@ -22,7 +22,6 @@
>   *   must also be changed
>   */
>  struct thread_info {
> -       struct task_struct      *task;          /* main task structure */
>         unsigned long           flags;          /* low level flags */
>         unsigned long           tp_value;       /* thread pointer */
>         __u32                   cpu;            /* current CPU */
> @@ -37,20 +36,11 @@ struct thread_info {
>   */
>  #define INIT_THREAD_INFO(tsk)                  \
>  {                                              \
> -       .task           = &tsk,                 \
>         .flags          = _TIF_FIXADE,          \
>         .cpu            = 0,                    \
>         .preempt_count  = INIT_PREEMPT_COUNT,   \
>  }
>
> -/* How to get the thread information struct from C. */
> -register struct thread_info *__current_thread_info __asm__("$tp");
> -
> -static inline struct thread_info *current_thread_info(void)
> -{
> -       return __current_thread_info;
> -}
> -
>  register unsigned long current_stack_pointer __asm__("$sp");
>
>  #endif /* !__ASSEMBLER__ */
> diff --git a/arch/loongarch/kernel/asm-offsets.c b/arch/loongarch/kernel/asm-offsets.c
> index 2cc953f113ac..453b176fa2d1 100644
> --- a/arch/loongarch/kernel/asm-offsets.c
> +++ b/arch/loongarch/kernel/asm-offsets.c
> @@ -70,7 +70,7 @@ static void __used output_task_defines(void)
>  {
>         COMMENT("LoongArch task_struct offsets.");
>         OFFSET(TASK_STATE, task_struct, __state);
> -       OFFSET(TASK_THREAD_INFO, task_struct, stack);
> +       OFFSET(TASK_STACK, task_struct, stack);
>         OFFSET(TASK_FLAGS, task_struct, flags);
>         OFFSET(TASK_MM, task_struct, mm);
>         OFFSET(TASK_PID, task_struct, pid);
> @@ -84,7 +84,6 @@ static void __used output_task_defines(void)
>  static void __used output_thread_info_defines(void)
>  {
>         COMMENT("LoongArch thread_info offsets.");
> -       OFFSET(TI_TASK, thread_info, task);
>         OFFSET(TI_FLAGS, thread_info, flags);
>         OFFSET(TI_TP_VALUE, thread_info, tp_value);
>         OFFSET(TI_CPU, thread_info, cpu);
> @@ -267,7 +266,8 @@ static void __used output_smpboot_defines(void)
>  {
>         COMMENT("Linux smp cpu boot offsets.");
>         OFFSET(CPU_BOOT_STACK, secondary_data, stack);
> -       OFFSET(CPU_BOOT_TINFO, secondary_data, thread_info);
> +       OFFSET(CPU_BOOT_TASK, secondary_data, task);
> +       OFFSET(CPU_BOOT_OFFSET, secondary_data, offset);
>         BLANK();
>  }
>  #endif
> diff --git a/arch/loongarch/kernel/entry.S b/arch/loongarch/kernel/entry.S
> index b53d333a7c42..53bce27f516b 100644
> --- a/arch/loongarch/kernel/entry.S
> +++ b/arch/loongarch/kernel/entry.S
> @@ -67,8 +67,11 @@ SYM_CODE_START(handle_syscall)
>  #endif
>
>         move            u0, t0
> -       LONG_LI         tp, ~_THREAD_MASK
> -       and             tp, tp, sp
> +       la_abs          t1, cpu_tasks
> +#ifdef CONFIG_SMP
> +       LONG_ADD        t1, t1, u0
> +#endif
> +       LONG_L          tp, t1, 0
>
>         move            a0, sp
>         bl              do_syscall
> diff --git a/arch/loongarch/kernel/head.S b/arch/loongarch/kernel/head.S
> index 4eed7bc312a8..ec67faab907b 100644
> --- a/arch/loongarch/kernel/head.S
> +++ b/arch/loongarch/kernel/head.S
> @@ -74,10 +74,11 @@ SYM_CODE_START(kernel_entry)                        # kernel entry point
>         /* GPR21 used for percpu base (runtime), initialized as 0 */
>         move            u0, zero
>
> -       la.pcrel        tp, init_thread_union
> -       /* Set the SP after an empty pt_regs.  */
> -       PTR_LI          sp, (_THREAD_SIZE - PT_SIZE)
> -       PTR_ADD         sp, sp, tp
> +       la.pcrel        tp, init_task
> +       la.pcrel        t0, init_stack
> +       PTR_LI          t1, _THREAD_SIZE
> +       PTR_ADD         t0, t0, t1
> +       PTR_ADDI        sp, t0, -PT_SIZE
>         set_saved_sp    sp, t0, t1
>
>  #ifdef CONFIG_RELOCATABLE
> @@ -86,8 +87,10 @@ SYM_CODE_START(kernel_entry)                 # kernel entry point
>
>  #ifdef CONFIG_RANDOMIZE_BASE
>         /* Repoint the sp into the new kernel */
> -       PTR_LI          sp, (_THREAD_SIZE - PT_SIZE)
> -       PTR_ADD         sp, sp, tp
> +       LONG_LPTR       t0, tp, TASK_STACK
> +       PTR_LI          t1, _THREAD_SIZE
> +       PTR_ADD         t0, t0, t1
> +       PTR_ADDI        sp, t0, -PT_SIZE
>         set_saved_sp    sp, t0, t1
>
>         /* Jump to the new kernel: new_pc = current_pc + random_offset */
> @@ -128,7 +131,8 @@ SYM_CODE_START(smpboot_entry)
>  #endif
>         la.pcrel        t0, cpuboot_data
>         ld.d            sp, t0, CPU_BOOT_STACK
> -       ld.d            tp, t0, CPU_BOOT_TINFO
> +       ld.d            tp, t0, CPU_BOOT_TASK
> +       ld.d            u0, t0, CPU_BOOT_OFFSET
>
>         bl              start_secondary
>         ASM_BUG()
> diff --git a/arch/loongarch/kernel/process.c b/arch/loongarch/kernel/process.c
> index 5505fc355e1b..de8b3cfa859e 100644
> --- a/arch/loongarch/kernel/process.c
> +++ b/arch/loongarch/kernel/process.c
> @@ -54,6 +54,9 @@
>  #include <asm/vdso.h>
>  #include <asm/vdso/vdso.h>
>
> +DEFINE_PER_CPU(struct task_struct *, cpu_tasks);
> +EXPORT_PER_CPU_SYMBOL_GPL(cpu_tasks);
I'm not sure but I think it is unnecessary to export.

> +
>  #ifdef CONFIG_STACKPROTECTOR
>  #include <linux/stackprotector.h>
>  unsigned long __stack_chk_guard __read_mostly;
> diff --git a/arch/loongarch/kernel/relocate.c b/arch/loongarch/kernel/relocate.c
> index 4b61a9632a98..2a42874e5eb7 100644
> --- a/arch/loongarch/kernel/relocate.c
> +++ b/arch/loongarch/kernel/relocate.c
> @@ -313,7 +313,7 @@ unsigned long __init relocate_kernel(void)
>                 reloc_offset += random_offset;
>
>                 /* The current thread is now within the relocated kernel */
> -               __current_thread_info = RELOCATED_KASLR(__current_thread_info);
> +               current_thread_pointer = RELOCATED_KASLR(current_thread_pointer);
>
>                 update_reloc_offset(&reloc_offset, random_offset);
>         }
> diff --git a/arch/loongarch/kernel/setup.c b/arch/loongarch/kernel/setup.c
> index 839b23edee87..e9618ceefc4a 100644
> --- a/arch/loongarch/kernel/setup.c
> +++ b/arch/loongarch/kernel/setup.c
> @@ -593,6 +593,7 @@ void __init setup_arch(char **cmdline_p)
>  {
>         cpu_probe();
>         unwind_init();
> +       set_current(&init_task);
You can also use set_current(current) here, it is a little faster,
because init_task has already been loaded in TP.

>
>         init_environ();
>         efi_init();
> diff --git a/arch/loongarch/kernel/smp.c b/arch/loongarch/kernel/smp.c
> index 64a048f1b880..fd7b4abf0f3e 100644
> --- a/arch/loongarch/kernel/smp.c
> +++ b/arch/loongarch/kernel/smp.c
> @@ -400,8 +400,9 @@ void loongson_boot_secondary(int cpu, struct task_struct *idle)
>         pr_info("Booting CPU#%d...\n", cpu);
>
>         entry = __pa_symbol((unsigned long)&smpboot_entry);
> -       cpuboot_data.stack = (unsigned long)__KSTK_TOS(idle);
> -       cpuboot_data.thread_info = (unsigned long)task_thread_info(idle);
> +       cpuboot_data.stack = (unsigned long)task_pt_regs(idle);
> +       cpuboot_data.task = (unsigned long)idle;
> +       cpuboot_data.offset = per_cpu_offset(cpu);
>
>         csr_mail_send(entry, cpu_logical_map(cpu), 0);
>
> @@ -663,6 +664,7 @@ asmlinkage void start_secondary(void)
>         set_my_cpu_offset(per_cpu_offset(cpu));
>
>         cpu_probe();
> +       set_current(current);
>         constant_clockevent_init();
>         loongson_init_secondary();
>
> diff --git a/arch/loongarch/kernel/switch.S b/arch/loongarch/kernel/switch.S
> index f377d8f5c51a..33a10221d73a 100644
> --- a/arch/loongarch/kernel/switch.S
> +++ b/arch/loongarch/kernel/switch.S
> @@ -12,7 +12,7 @@
>
>  /*
>   * task_struct *__switch_to(task_struct *prev, task_struct *next,
> - *                         struct thread_info *next_ti, void *sched_ra, void *sched_cfa)
> + *                         void *sched_ra, void *sched_cfa)
>   */
>         .align  5
>  SYM_FUNC_START(__switch_to)
> @@ -24,8 +24,8 @@ SYM_FUNC_START(__switch_to)
>         LONG_SPTR       t1, a0, (THREAD_CSRPRMD - TASK_STRUCT_OFFSET)
>
>         cpu_save_nonscratch a0
> -       LONG_SPTR       a3, a0, (THREAD_SCHED_RA - TASK_STRUCT_OFFSET)
> -       LONG_SPTR       a4, a0, (THREAD_SCHED_CFA - TASK_STRUCT_OFFSET)
> +       LONG_SPTR       a2, a0, (THREAD_SCHED_RA - TASK_STRUCT_OFFSET)
> +       LONG_SPTR       a3, a0, (THREAD_SCHED_CFA - TASK_STRUCT_OFFSET)
>
>  #if defined(CONFIG_STACKPROTECTOR) && !defined(CONFIG_SMP)
>         la              t7, __stack_chk_guard
> @@ -33,11 +33,12 @@ SYM_FUNC_START(__switch_to)
>         LONG_SPTR       t8, t7, 0
>  #endif
>
> -       move    tp, a2
> +       move    tp, a1
>         cpu_restore_nonscratch a1
>
> -       li.w            t0, _THREAD_SIZE
> -       PTR_ADD         t0, t0, tp
> +       LONG_LPTR       t0, tp, (TASK_STACK - TASK_STRUCT_OFFSET)
> +       PTR_LI          t1, _THREAD_SIZE
> +       PTR_ADD         t0, t0, t1
>         set_saved_sp    t0, t1, t2
>
>         LONG_LPTR       t1, a1, (THREAD_CSRPRMD - TASK_STRUCT_OFFSET)
As AI said, TP of the next task should be adjusted with offset, the
best way is adding "PTR_ADDI tp, a2, -TASK_STRUCT_OFFSET" in the last
ifdefs.


Huacai

> --
> 2.42.0
>