| Message ID | 1416977391-24231-3-git-send-email-takahiro.akashi@linaro.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Wed, Nov 26, 2014 at 04:49:47AM +0000, AKASHI Takahiro wrote: > If tracer modifies a syscall number to -1, this traced system call should > be skipped with a return value specified in x0. > This patch implements this semantics. > > Please note: > * syscall entry tracing and syscall exit tracing (ftrace tracepoint and > audit) are always executed, if enabled, even when skipping a system call > (that is, -1). > In this way, we can avoid a potential bug where audit_syscall_entry() > might be called without audit_syscall_exit() at the previous system call > being called, that would cause OOPs in audit_syscall_entry(). > > Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org> > --- > arch/arm64/kernel/entry.S | 10 +++++++++- > 1 file changed, 9 insertions(+), 1 deletion(-) > > diff --git a/arch/arm64/kernel/entry.S b/arch/arm64/kernel/entry.S > index 726b910..946ec52 100644 > --- a/arch/arm64/kernel/entry.S > +++ b/arch/arm64/kernel/entry.S > @@ -161,6 +161,7 @@ > */ > sc_nr .req x25 // number of system calls > scno .req x26 // syscall number > +scno_w .req w26 // syscall number (lower 32 bits) > stbl .req x27 // syscall table pointer > tsk .req x28 // current thread_info > > @@ -668,8 +669,14 @@ ENDPROC(el0_svc) > * switches, and waiting for our parent to respond. > */ > __sys_trace: > - mov x0, sp > + cmp scno_w, #-1 // set default errno for I hate that we have to use scno_w, but the only alternative I can think of is using w8 directly, which isn't any better and doesn't work for compat. Ho-hum, I guess we'll stick with what you have. > + b.ne 1f // user-issued syscall(-1) > + mov x0, #-ENOSYS > + str x0, [sp] Can you use #S_X0 here for clarity, please? > +1: mov x0, sp > bl syscall_trace_enter > + cmp w0, #-1 // skip the syscall? > + b.eq __sys_trace_return_skipped > adr lr, __sys_trace_return // return address > uxtw scno, w0 // syscall number (possibly new) > mov x1, sp // pointer to regs > @@ -684,6 +691,7 @@ __sys_trace: > > __sys_trace_return: > str x0, [sp] // save returned x0 and update this guy too. Will
On 11/26/2014 10:02 PM, Will Deacon wrote: > On Wed, Nov 26, 2014 at 04:49:47AM +0000, AKASHI Takahiro wrote: >> If tracer modifies a syscall number to -1, this traced system call should >> be skipped with a return value specified in x0. >> This patch implements this semantics. >> >> Please note: >> * syscall entry tracing and syscall exit tracing (ftrace tracepoint and >> audit) are always executed, if enabled, even when skipping a system call >> (that is, -1). >> In this way, we can avoid a potential bug where audit_syscall_entry() >> might be called without audit_syscall_exit() at the previous system call >> being called, that would cause OOPs in audit_syscall_entry(). >> >> Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org> >> --- >> arch/arm64/kernel/entry.S | 10 +++++++++- >> 1 file changed, 9 insertions(+), 1 deletion(-) >> >> diff --git a/arch/arm64/kernel/entry.S b/arch/arm64/kernel/entry.S >> index 726b910..946ec52 100644 >> --- a/arch/arm64/kernel/entry.S >> +++ b/arch/arm64/kernel/entry.S >> @@ -161,6 +161,7 @@ >> */ >> sc_nr .req x25 // number of system calls >> scno .req x26 // syscall number >> +scno_w .req w26 // syscall number (lower 32 bits) >> stbl .req x27 // syscall table pointer >> tsk .req x28 // current thread_info >> >> @@ -668,8 +669,14 @@ ENDPROC(el0_svc) >> * switches, and waiting for our parent to respond. >> */ >> __sys_trace: >> - mov x0, sp >> + cmp scno_w, #-1 // set default errno for > > I hate that we have to use scno_w, but the only alternative I can think of > is using w8 directly, which isn't any better and doesn't work for compat. > Ho-hum, I guess we'll stick with what you have. The possible approaches might be: * use 32-bit registers for scno & sc_nr and use "sxtw scno, w8," or * use an extra reg like __sys_trace: mov x0, #0xffffffff cmp scno, x0 b.ne 1f >> + b.ne 1f // user-issued syscall(-1) >> + mov x0, #-ENOSYS >> + str x0, [sp] > > Can you use #S_X0 here for clarity, please? Okey. >> +1: mov x0, sp >> bl syscall_trace_enter >> + cmp w0, #-1 // skip the syscall? >> + b.eq __sys_trace_return_skipped >> adr lr, __sys_trace_return // return address >> uxtw scno, w0 // syscall number (possibly new) >> mov x1, sp // pointer to regs >> @@ -684,6 +691,7 @@ __sys_trace: >> >> __sys_trace_return: >> str x0, [sp] // save returned x0 > > and update this guy too. Sure. -Takahiro AKASHI > Will > -- > To unsubscribe from this list: send the line "unsubscribe linux-kernel" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html > Please read the FAQ at http://www.tux.org/lkml/ >
On Thu, Nov 27, 2014 at 06:46:29AM +0000, AKASHI Takahiro wrote: > On 11/26/2014 10:02 PM, Will Deacon wrote: > > On Wed, Nov 26, 2014 at 04:49:47AM +0000, AKASHI Takahiro wrote: > >> If tracer modifies a syscall number to -1, this traced system call should > >> be skipped with a return value specified in x0. > >> This patch implements this semantics. > >> > >> Please note: > >> * syscall entry tracing and syscall exit tracing (ftrace tracepoint and > >> audit) are always executed, if enabled, even when skipping a system call > >> (that is, -1). > >> In this way, we can avoid a potential bug where audit_syscall_entry() > >> might be called without audit_syscall_exit() at the previous system call > >> being called, that would cause OOPs in audit_syscall_entry(). > >> > >> Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org> > >> --- > >> arch/arm64/kernel/entry.S | 10 +++++++++- > >> 1 file changed, 9 insertions(+), 1 deletion(-) > >> > >> diff --git a/arch/arm64/kernel/entry.S b/arch/arm64/kernel/entry.S > >> index 726b910..946ec52 100644 > >> --- a/arch/arm64/kernel/entry.S > >> +++ b/arch/arm64/kernel/entry.S > >> @@ -161,6 +161,7 @@ > >> */ > >> sc_nr .req x25 // number of system calls > >> scno .req x26 // syscall number > >> +scno_w .req w26 // syscall number (lower 32 bits) > >> stbl .req x27 // syscall table pointer > >> tsk .req x28 // current thread_info > >> > >> @@ -668,8 +669,14 @@ ENDPROC(el0_svc) > >> * switches, and waiting for our parent to respond. > >> */ > >> __sys_trace: > >> - mov x0, sp > >> + cmp scno_w, #-1 // set default errno for > > > > I hate that we have to use scno_w, but the only alternative I can think of > > is using w8 directly, which isn't any better and doesn't work for compat. > > Ho-hum, I guess we'll stick with what you have. > > The possible approaches might be: > * use 32-bit registers for scno & sc_nr and use "sxtw scno, w8," or > * use an extra reg like > __sys_trace: > mov x0, #0xffffffff or even better: mov w0, #-1 > cmp scno, x0 > b.ne 1f Good thinking, I prefer that. Will
diff --git a/arch/arm64/kernel/entry.S b/arch/arm64/kernel/entry.S index 726b910..946ec52 100644 --- a/arch/arm64/kernel/entry.S +++ b/arch/arm64/kernel/entry.S @@ -161,6 +161,7 @@ */ sc_nr .req x25 // number of system calls scno .req x26 // syscall number +scno_w .req w26 // syscall number (lower 32 bits) stbl .req x27 // syscall table pointer tsk .req x28 // current thread_info @@ -668,8 +669,14 @@ ENDPROC(el0_svc) * switches, and waiting for our parent to respond. */ __sys_trace: - mov x0, sp + cmp scno_w, #-1 // set default errno for + b.ne 1f // user-issued syscall(-1) + mov x0, #-ENOSYS + str x0, [sp] +1: mov x0, sp bl syscall_trace_enter + cmp w0, #-1 // skip the syscall? + b.eq __sys_trace_return_skipped adr lr, __sys_trace_return // return address uxtw scno, w0 // syscall number (possibly new) mov x1, sp // pointer to regs @@ -684,6 +691,7 @@ __sys_trace: __sys_trace_return: str x0, [sp] // save returned x0 +__sys_trace_return_skipped: mov x0, sp bl syscall_trace_exit b ret_to_user
If tracer modifies a syscall number to -1, this traced system call should be skipped with a return value specified in x0. This patch implements this semantics. Please note: * syscall entry tracing and syscall exit tracing (ftrace tracepoint and audit) are always executed, if enabled, even when skipping a system call (that is, -1). In this way, we can avoid a potential bug where audit_syscall_entry() might be called without audit_syscall_exit() at the previous system call being called, that would cause OOPs in audit_syscall_entry(). Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org> --- arch/arm64/kernel/entry.S | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-)