[PATCH V3 3/9] fork: move PF_IO_WORKER's kernel frame setup to new flag

From: Mike Christie
Date: Mon Oct 04 2021 - 15:22:36 EST


The vhost worker threads need the same frame setup as io_uring's worker
threads, but handle signals differently and do not need the same
scheduling behavior. This patch separate's the frame setup parts of
PF_IO_WORKER into a kernel_clone_args flag, KERN_WORKER_USER.

Signed-off-by: Mike Christie <michael.christie@xxxxxxxxxx>
Acked-by: Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>
Acked-by: Christian Brauner <christian.brauner@xxxxxxxxxx>
---
arch/alpha/kernel/process.c | 3 ++-
arch/arc/kernel/process.c | 3 ++-
arch/arm/kernel/process.c | 3 ++-
arch/arm64/kernel/process.c | 3 ++-
arch/csky/kernel/process.c | 3 ++-
arch/h8300/kernel/process.c | 3 ++-
arch/hexagon/kernel/process.c | 3 ++-
arch/ia64/kernel/process.c | 3 ++-
arch/m68k/kernel/process.c | 3 ++-
arch/microblaze/kernel/process.c | 3 ++-
arch/mips/kernel/process.c | 3 ++-
arch/nds32/kernel/process.c | 3 ++-
arch/nios2/kernel/process.c | 3 ++-
arch/openrisc/kernel/process.c | 3 ++-
arch/parisc/kernel/process.c | 3 ++-
arch/powerpc/kernel/process.c | 3 ++-
arch/riscv/kernel/process.c | 3 ++-
arch/s390/kernel/process.c | 3 ++-
arch/sh/kernel/process_32.c | 3 ++-
arch/sparc/kernel/process_32.c | 3 ++-
arch/sparc/kernel/process_64.c | 3 ++-
arch/um/kernel/process.c | 3 ++-
arch/x86/kernel/process.c | 4 ++--
arch/xtensa/kernel/process.c | 3 ++-
include/linux/sched/task.h | 1 +
kernel/fork.c | 2 +-
26 files changed, 50 insertions(+), 26 deletions(-)

diff --git a/arch/alpha/kernel/process.c b/arch/alpha/kernel/process.c
index 6005b0dfe7e2..e9b2dde444f4 100644
--- a/arch/alpha/kernel/process.c
+++ b/arch/alpha/kernel/process.c
@@ -249,7 +249,8 @@ int copy_thread(unsigned long clone_flags, unsigned long usp,
childti->pcb.ksp = (unsigned long) childstack;
childti->pcb.flags = 1; /* set FEN, clear everything else */

- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
/* kernel thread */
memset(childstack, 0,
sizeof(struct switch_stack) + sizeof(struct pt_regs));
diff --git a/arch/arc/kernel/process.c b/arch/arc/kernel/process.c
index 4e307e5b5205..2caa80fb9e9c 100644
--- a/arch/arc/kernel/process.c
+++ b/arch/arc/kernel/process.c
@@ -191,7 +191,8 @@ int copy_thread(unsigned long clone_flags, unsigned long usp,
childksp[0] = 0; /* fp */
childksp[1] = (unsigned long)ret_from_fork; /* blink */

- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
memset(c_regs, 0, sizeof(struct pt_regs));

c_callee->r13 = kthread_arg;
diff --git a/arch/arm/kernel/process.c b/arch/arm/kernel/process.c
index 07ae4444b6ab..9f41435d78d9 100644
--- a/arch/arm/kernel/process.c
+++ b/arch/arm/kernel/process.c
@@ -248,7 +248,8 @@ int copy_thread(unsigned long clone_flags, unsigned long stack_start,
thread->cpu_domain = get_domain();
#endif

- if (likely(!(p->flags & (PF_KTHREAD | PF_IO_WORKER)))) {
+ if (likely(!((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER)))) {
*childregs = *current_pt_regs();
childregs->ARM_r0 = 0;
if (stack_start)
diff --git a/arch/arm64/kernel/process.c b/arch/arm64/kernel/process.c
index 7979ec253c29..d149de03bd50 100644
--- a/arch/arm64/kernel/process.c
+++ b/arch/arm64/kernel/process.c
@@ -334,7 +334,8 @@ int copy_thread(unsigned long clone_flags, unsigned long stack_start,

ptrauth_thread_init_kernel(p);

- if (likely(!(p->flags & (PF_KTHREAD | PF_IO_WORKER)))) {
+ if (likely(!((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER)))) {
*childregs = *current_pt_regs();
childregs->regs[0] = 0;

diff --git a/arch/csky/kernel/process.c b/arch/csky/kernel/process.c
index f38b668515ae..10debf43375e 100644
--- a/arch/csky/kernel/process.c
+++ b/arch/csky/kernel/process.c
@@ -50,7 +50,8 @@ int copy_thread(unsigned long clone_flags,
/* setup thread.sp for switch_to !!! */
p->thread.sp = (unsigned long)childstack;

- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
memset(childregs, 0, sizeof(struct pt_regs));
childstack->r15 = (unsigned long) ret_from_kernel_thread;
childstack->r10 = kthread_arg;
diff --git a/arch/h8300/kernel/process.c b/arch/h8300/kernel/process.c
index 9a8f6c033ad1..e0d69c3afa2a 100644
--- a/arch/h8300/kernel/process.c
+++ b/arch/h8300/kernel/process.c
@@ -113,7 +113,8 @@ int copy_thread(unsigned long clone_flags, unsigned long usp,

childregs = (struct pt_regs *) (THREAD_SIZE + task_stack_page(p)) - 1;

- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
memset(childregs, 0, sizeof(struct pt_regs));
childregs->retpc = (unsigned long) ret_from_kernel_thread;
childregs->er4 = topstk; /* arg */
diff --git a/arch/hexagon/kernel/process.c b/arch/hexagon/kernel/process.c
index 664367be55e5..9ea473567a5c 100644
--- a/arch/hexagon/kernel/process.c
+++ b/arch/hexagon/kernel/process.c
@@ -73,7 +73,8 @@ int copy_thread(unsigned long clone_flags, unsigned long usp, unsigned long arg,
sizeof(*ss));
ss->lr = (unsigned long)ret_from_fork;
p->thread.switch_sp = ss;
- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
memset(childregs, 0, sizeof(struct pt_regs));
/* r24 <- fn, r25 <- arg */
ss->r24 = usp;
diff --git a/arch/ia64/kernel/process.c b/arch/ia64/kernel/process.c
index a69cc33b5e32..d7c47ea12703 100644
--- a/arch/ia64/kernel/process.c
+++ b/arch/ia64/kernel/process.c
@@ -339,7 +339,8 @@ copy_thread(unsigned long clone_flags, unsigned long user_stack_base,

ia64_drop_fpu(p); /* don't pick up stale state from a CPU's fph */

- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
if (unlikely(!user_stack_base)) {
/* fork_idle() called us */
return 0;
diff --git a/arch/m68k/kernel/process.c b/arch/m68k/kernel/process.c
index 7587291793fb..a842e6c7bef2 100644
--- a/arch/m68k/kernel/process.c
+++ b/arch/m68k/kernel/process.c
@@ -157,7 +157,8 @@ int copy_thread(unsigned long clone_flags, unsigned long usp, unsigned long arg,
*/
p->thread.fc = USER_DATA;

- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
/* kernel thread */
memset(frame, 0, sizeof(struct fork_frame));
frame->regs.sr = PS_S;
diff --git a/arch/microblaze/kernel/process.c b/arch/microblaze/kernel/process.c
index b8eb544e1fd6..ba1a45842a70 100644
--- a/arch/microblaze/kernel/process.c
+++ b/arch/microblaze/kernel/process.c
@@ -59,7 +59,8 @@ int copy_thread(unsigned long clone_flags, unsigned long usp, unsigned long arg,
struct pt_regs *childregs = task_pt_regs(p);
struct thread_info *ti = task_thread_info(p);

- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
/* if we're creating a new kernel thread then just zeroing all
* the registers. That's OK for a brand new thread.*/
memset(childregs, 0, sizeof(struct pt_regs));
diff --git a/arch/mips/kernel/process.c b/arch/mips/kernel/process.c
index d494e1d76e71..b592d1bfab09 100644
--- a/arch/mips/kernel/process.c
+++ b/arch/mips/kernel/process.c
@@ -120,7 +120,8 @@ int copy_thread(unsigned long clone_flags, unsigned long usp,
/* Put the stack after the struct pt_regs. */
childksp = (unsigned long) childregs;
p->thread.cp0_status = (read_c0_status() & ~(ST0_CU2|ST0_CU1)) | ST0_KERNEL_CUMASK;
- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
/* kernel thread */
unsigned long status = p->thread.cp0_status;
memset(childregs, 0, sizeof(struct pt_regs));
diff --git a/arch/nds32/kernel/process.c b/arch/nds32/kernel/process.c
index 1ca8900f9d07..8ec5b725842b 100644
--- a/arch/nds32/kernel/process.c
+++ b/arch/nds32/kernel/process.c
@@ -157,7 +157,8 @@ int copy_thread(unsigned long clone_flags, unsigned long stack_start,

memset(&p->thread.cpu_context, 0, sizeof(struct cpu_context));

- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
memset(childregs, 0, sizeof(struct pt_regs));
/* kernel thread fn */
p->thread.cpu_context.r6 = stack_start;
diff --git a/arch/nios2/kernel/process.c b/arch/nios2/kernel/process.c
index b49dc6500118..e22b83b64769 100644
--- a/arch/nios2/kernel/process.c
+++ b/arch/nios2/kernel/process.c
@@ -109,7 +109,8 @@ int copy_thread(unsigned long clone_flags, unsigned long usp, unsigned long arg,
struct switch_stack *childstack =
((struct switch_stack *)childregs) - 1;

- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
memset(childstack, 0,
sizeof(struct switch_stack) + sizeof(struct pt_regs));

diff --git a/arch/openrisc/kernel/process.c b/arch/openrisc/kernel/process.c
index 7b356a9a8dc7..684ef1f0999c 100644
--- a/arch/openrisc/kernel/process.c
+++ b/arch/openrisc/kernel/process.c
@@ -173,7 +173,8 @@ copy_thread(unsigned long clone_flags, unsigned long usp, unsigned long arg,
sp -= sizeof(struct pt_regs);
kregs = (struct pt_regs *)sp;

- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
memset(kregs, 0, sizeof(struct pt_regs));
kregs->gpr[20] = usp; /* fn, kernel thread */
kregs->gpr[22] = arg;
diff --git a/arch/parisc/kernel/process.c b/arch/parisc/kernel/process.c
index d9555ccf1e9c..1c955e6bad83 100644
--- a/arch/parisc/kernel/process.c
+++ b/arch/parisc/kernel/process.c
@@ -198,7 +198,8 @@ copy_thread(unsigned long clone_flags, unsigned long usp,
extern void * const ret_from_kernel_thread;
extern void * const child_return;

- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
/* kernel thread */
memset(cregs, 0, sizeof(struct pt_regs));
if (!usp) /* idle thread */
diff --git a/arch/powerpc/kernel/process.c b/arch/powerpc/kernel/process.c
index d2f2301b0ad1..097f13b43a8f 100644
--- a/arch/powerpc/kernel/process.c
+++ b/arch/powerpc/kernel/process.c
@@ -1700,7 +1700,8 @@ int copy_thread(unsigned long clone_flags, unsigned long usp,
/* Copy registers */
sp -= sizeof(struct pt_regs);
childregs = (struct pt_regs *) sp;
- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
/* kernel thread */
memset(childregs, 0, sizeof(struct pt_regs));
childregs->gpr[1] = sp + sizeof(struct pt_regs);
diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c
index 3d0e6390f34c..39bb4a79be15 100644
--- a/arch/riscv/kernel/process.c
+++ b/arch/riscv/kernel/process.c
@@ -125,7 +125,8 @@ int copy_thread(unsigned long clone_flags, unsigned long usp, unsigned long arg,
struct pt_regs *childregs = task_pt_regs(p);

/* p->thread holds context to be restored by __switch_to() */
- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
/* Kernel thread */
memset(childregs, 0, sizeof(struct pt_regs));
childregs->gp = gp_in_global;
diff --git a/arch/s390/kernel/process.c b/arch/s390/kernel/process.c
index 01b969bb868e..29ba92911340 100644
--- a/arch/s390/kernel/process.c
+++ b/arch/s390/kernel/process.c
@@ -131,7 +131,8 @@ int copy_thread(unsigned long clone_flags, unsigned long new_stackp,
frame->sf.gprs[9] = (unsigned long)frame;

/* Store access registers to kernel stack of new process. */
- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
/* kernel thread */
memset(&frame->childregs, 0, sizeof(struct pt_regs));
frame->childregs.psw.mask = PSW_KERNEL_BITS | PSW_MASK_DAT |
diff --git a/arch/sh/kernel/process_32.c b/arch/sh/kernel/process_32.c
index d199805552c0..8cbd7404df40 100644
--- a/arch/sh/kernel/process_32.c
+++ b/arch/sh/kernel/process_32.c
@@ -114,7 +114,8 @@ int copy_thread(unsigned long clone_flags, unsigned long usp, unsigned long arg,

childregs = task_pt_regs(p);
p->thread.sp = (unsigned long) childregs;
- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
memset(childregs, 0, sizeof(struct pt_regs));
p->thread.pc = (unsigned long) ret_from_kernel_thread;
childregs->regs[4] = arg;
diff --git a/arch/sparc/kernel/process_32.c b/arch/sparc/kernel/process_32.c
index 6e04cfc64b99..2522283a63ac 100644
--- a/arch/sparc/kernel/process_32.c
+++ b/arch/sparc/kernel/process_32.c
@@ -296,7 +296,8 @@ int copy_thread(unsigned long clone_flags, unsigned long sp, unsigned long arg,
ti->ksp = (unsigned long) new_stack;
p->thread.kregs = childregs;

- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
extern int nwindows;
unsigned long psr;
memset(new_stack, 0, STACKFRAME_SZ + TRACEREG_SZ);
diff --git a/arch/sparc/kernel/process_64.c b/arch/sparc/kernel/process_64.c
index b339eaa1f890..a157474c970f 100644
--- a/arch/sparc/kernel/process_64.c
+++ b/arch/sparc/kernel/process_64.c
@@ -594,7 +594,8 @@ int copy_thread(unsigned long clone_flags, unsigned long sp, unsigned long arg,
sizeof(struct sparc_stackf));
t->fpsaved[0] = 0;

- if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (unlikely((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
memset(child_trap_frame, 0, child_stack_sz);
__thread_flag_byte_ptr(t)[TI_FLAG_BYTE_CWP] =
(current_pt_regs()->tstate + 1) & TSTATE_CWP;
diff --git a/arch/um/kernel/process.c b/arch/um/kernel/process.c
index 0815a43b9f4a..28e5c9f67436 100644
--- a/arch/um/kernel/process.c
+++ b/arch/um/kernel/process.c
@@ -158,7 +158,8 @@ int copy_thread(unsigned long clone_flags, unsigned long sp,
u32 worker_flags)
{
void (*handler)(void);
- int kthread = current->flags & (PF_KTHREAD | PF_IO_WORKER);
+ int kthread = (current->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER);
int ret = 0;

p->thread = (struct thread_struct) INIT_THREAD;
diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
index 20d9bab61b14..a904f5524d73 100644
--- a/arch/x86/kernel/process.c
+++ b/arch/x86/kernel/process.c
@@ -178,9 +178,9 @@ int copy_thread(unsigned long clone_flags, unsigned long sp, unsigned long arg,
task_user_gs(p) = get_user_gs(current_pt_regs());
#endif

- if (unlikely(p->flags & PF_IO_WORKER)) {
+ if (unlikely(worker_flags & KERN_WORKER_USER)) {
/*
- * An IO thread is a user space thread, but it doesn't
+ * A user worker thread is a user space thread, but it doesn't
* return to ret_after_fork().
*
* In order to indicate that to tools like gdb,
diff --git a/arch/xtensa/kernel/process.c b/arch/xtensa/kernel/process.c
index a0ad9f0cc0cf..0af51e94c8dc 100644
--- a/arch/xtensa/kernel/process.c
+++ b/arch/xtensa/kernel/process.c
@@ -217,7 +217,8 @@ int copy_thread(unsigned long clone_flags, unsigned long usp_thread_fn,

p->thread.sp = (unsigned long)childregs;

- if (!(p->flags & (PF_KTHREAD | PF_IO_WORKER))) {
+ if (!((p->flags & PF_KTHREAD) ||
+ (worker_flags & KERN_WORKER_USER))) {
struct pt_regs *regs = current_pt_regs();
unsigned long usp = usp_thread_fn ?
usp_thread_fn : regs->areg[1];
diff --git a/include/linux/sched/task.h b/include/linux/sched/task.h
index ffc7c6a384ad..cf7c9fffc839 100644
--- a/include/linux/sched/task.h
+++ b/include/linux/sched/task.h
@@ -19,6 +19,7 @@ struct css_set;
#define CLONE_LEGACY_FLAGS 0xffffffffULL

#define KERN_WORKER_IO BIT(0)
+#define KERN_WORKER_USER BIT(1)

struct kernel_clone_args {
u64 flags;
diff --git a/kernel/fork.c b/kernel/fork.c
index 3c3624786e4d..4b0e8257993b 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -2527,7 +2527,7 @@ struct task_struct *create_io_thread(int (*fn)(void *), void *arg, int node)
.exit_signal = (lower_32_bits(flags) & CSIGNAL),
.stack = (unsigned long)fn,
.stack_size = (unsigned long)arg,
- .worker_flags = KERN_WORKER_IO,
+ .worker_flags = KERN_WORKER_IO | KERN_WORKER_USER,
};

return copy_process(NULL, 0, node, &args);
--
2.25.1