| Message ID | d4a65be8d05d945885f53bc168fd85f08e72adf1.1601960644.git.thehajime@gmail.com |
|---|---|
| State | Not Applicable |
| Headers | show |
| Series | [RFC,v7,01/21] um: split build in kernel and host parts | expand |
On Tue, 2020-10-06 at 18:44 +0900, Hajime Tazaki wrote: > This commit introduces additional argument of previous task when > context switch happens. New SUBARCH can use the new information to > switch tasks in a subarch-specific manner. > > The patch is particularly required by nommu mode implemented as a > SUBARCH of UML. Would probably be good to already say why here? johannes
On Thu, 08 Oct 2020 00:25:06 +0900, Johannes Berg wrote: > > On Tue, 2020-10-06 at 18:44 +0900, Hajime Tazaki wrote: > > This commit introduces additional argument of previous task when > > context switch happens. New SUBARCH can use the new information to > > switch tasks in a subarch-specific manner. > > > > The patch is particularly required by nommu mode implemented as a > > SUBARCH of UML. > > Would probably be good to already say why here? Agree. The patch is particularly required by nommu mode because it uses previous task information to control the previous thread (e.g., down semaphore, terminate thread, clean up thread flags). Something like this ? -- Hajime
On Fri, 2020-10-09 at 10:24 +0900, Hajime Tazaki wrote: > On Thu, 08 Oct 2020 00:25:06 +0900, > Johannes Berg wrote: > > On Tue, 2020-10-06 at 18:44 +0900, Hajime Tazaki wrote: > > > This commit introduces additional argument of previous task when > > > context switch happens. New SUBARCH can use the new information to > > > switch tasks in a subarch-specific manner. > > > > > > The patch is particularly required by nommu mode implemented as a > > > SUBARCH of UML. > > > > Would probably be good to already say why here? > > Agree. > > The patch is particularly required by nommu mode because it uses > previous task information to control the previous thread (e.g., down > semaphore, terminate thread, clean up thread flags). > > Something like this ? Well, still not sure I like "nommu" to refer to "library mode", but I'm also not sure that you can understand that without the context of how threads actually work in library mode? Maybe even something vague like Having access to the previous thread will be required in library mode for it to implement thread switching correctly. would give some rationale? johannes
diff --git a/arch/um/kernel/process.c b/arch/um/kernel/process.c index 26b5e243d3fc..87a8cfa228ca 100644 --- a/arch/um/kernel/process.c +++ b/arch/um/kernel/process.c @@ -80,7 +80,7 @@ static inline void set_current(struct task_struct *task) { external_pid(), task }); } -extern void arch_switch_to(struct task_struct *to); +extern void arch_switch_to(struct task_struct *from, struct task_struct *to); void *__switch_to(struct task_struct *from, struct task_struct *to) { @@ -88,7 +88,7 @@ void *__switch_to(struct task_struct *from, struct task_struct *to) set_current(to); switch_threads(&from->thread.switch_buf, &to->thread.switch_buf); - arch_switch_to(current); + arch_switch_to(from, to); return current->thread.prev_sched; } @@ -145,7 +145,7 @@ void fork_handler(void) * arch_switch_to isn't needed. We could want to apply this to * improve performance. -bb */ - arch_switch_to(current); + arch_switch_to(NULL, current); current->thread.prev_sched = NULL; diff --git a/arch/x86/um/ptrace_32.c b/arch/x86/um/ptrace_32.c index 2497bac56066..0f184710d4ca 100644 --- a/arch/x86/um/ptrace_32.c +++ b/arch/x86/um/ptrace_32.c @@ -11,7 +11,7 @@ extern int arch_switch_tls(struct task_struct *to); -void arch_switch_to(struct task_struct *to) +void arch_switch_to(struct task_struct *from, struct task_struct *to) { int err = arch_switch_tls(to); if (!err) diff --git a/arch/x86/um/syscalls_64.c b/arch/x86/um/syscalls_64.c index 58f51667e2e4..2ef9474d2bd2 100644 --- a/arch/x86/um/syscalls_64.c +++ b/arch/x86/um/syscalls_64.c @@ -80,7 +80,7 @@ SYSCALL_DEFINE2(arch_prctl, int, option, unsigned long, arg2) return arch_prctl(current, option, (unsigned long __user *) arg2); } -void arch_switch_to(struct task_struct *to) +void arch_switch_to(struct task_struct *from, struct task_struct *to) { if ((to->thread.arch.fs == 0) || (to->mm == NULL)) return;
This commit introduces additional argument of previous task when context switch happens. New SUBARCH can use the new information to switch tasks in a subarch-specific manner. The patch is particularly required by nommu mode implemented as a SUBARCH of UML. Signed-off-by: Hajime Tazaki <thehajime@gmail.com> --- arch/um/kernel/process.c | 6 +++--- arch/x86/um/ptrace_32.c | 2 +- arch/x86/um/syscalls_64.c | 2 +- 3 files changed, 5 insertions(+), 5 deletions(-)