sparc64: Jump to boot prom from console on panic
Subsystems (4)
v3 cover sparc64: Jump to boot prom from console on panic
V3 changes:
- patch 02/04: Added SERIAL_SUNHV conditional group for sunhv_migrate_hvcons_irq in smp_send_stop().
V2 changes:
- Added cover letter patch
Hi,
Currently Stop-A (L1A) does not make the kernel switch to OBP on panic. This patchset addresses this issue. Also, now we can cause a jump to OBP by sending 'break' twice from sunhv console. On bare metal, one can send a break by typing Esc + 'B' + Sysrq (or whatever). On LDOM, press Ctrl + ] in telnet, and then "send break" at the telnet prompt.
Thanks.
sparc64: Set cpu state to offline when stopped
sparc64: Migrate hvcons irq to panicked cpu
sparc64: Send break twice from console to return to boot prom Documentation/sparc: Steps for sending break on sunhv console
Documentation/sparc/console.txt | 9 +++++++++ arch/sparc/include/asm/setup.h | 1 + arch/sparc/kernel/smp_64.c | 9 ++++++++- drivers/tty/serial/sunhv.c | 12 +++++++++++- kernel/panic.c | 3 ++- 5 files changed, 31 insertions(+), 3 deletions(-) create mode 100644 Documentation/sparc/console.txt
--
1.7.1
PATCH v3 1/4 sparc64: Set cpu state to offline when stopped
CPU needs to be marked offline before stopping it. When not marked offline, the xcall receives HV_EWOULDBLOCK and so assumes that not all CPUs received the message, and retries. After 10000 retries, it finally fails with fatal mondo timeout.
--- arch/sparc/kernel/smp_64.c | 3 +++ 1 files changed, 3 insertions(+), 0 deletions(-)
arch/sparc/kernel/smp_64.c +3 -0
@@ -1443,6 +1443,7 @@ void __irq_entry smp_receive_signal_client(int irq, struct pt_regs *regs)static void stop_this_cpu (void *dummy) { +set_cpu_online (smp_processor_id (), false);prom_stopself (); }
@@ -1454,6 +1455,8 @@ void smp_send_stop(void)for_each_online_cpu (cpu) {if (cpu =smp_processor_id ())continue ; + +set_cpu_online (cpu, false);#ifdef CONFIG_SUN_LDOMS if (ldom_domaining_enabled) {unsigned long hv_err;
Signature · 1 line
1.7.1
PATCH v3 2/4 sparc64: Migrate hvcons irq to panicked cpu
In reply to Vijay KumarOn panic, all other CPUs are stopped except the one which had hit panic. To keep console alive, we need to migrate hvcons irq to panicked CPU.
---v2->v3: Added SERIAL_SUNHV conditional group for
sunhv_migrate_hvcons_irq().
--- arch/sparc/include/asm/setup.h | 5 ++++- arch/sparc/kernel/smp_64.c | 6 +++++- drivers/tty/serial/sunhv.c | 6 ++++++ 3 files changed, 15 insertions(+), 2 deletions(-)
arch/sparc/include/asm/setup.h +4 -1
@@ -59,8 +59,11 @@ static inline int con_is_present(void)extern atomic_t dcpage_flushes_xcall;extern int sysctl_tsb_ratio; -#endif +#ifdef CONFIG_SERIAL_SUNHV +void sunhv_migrate_hvcons_irq (int cpu); +#endif +#endif void sun_do_break (void );extern int stop_a_enabled;extern int scons_pwroff;
arch/sparc/kernel/smp_64.c +5 -1
@@ -1452,8 +1452,12 @@ void smp_send_stop(void)int cpu;if (tlb_type = hypervisor) { +int this_cpu =smp_processor_id (); +#ifdef CONFIG_SERIAL_SUNHV +sunhv_migrate_hvcons_irq (this_cpu); +#endif for_each_online_cpu (cpu) { -if (cpu =smp_processor_id ()) +if (cpu = this_cpu)continue ;set_cpu_online (cpu, false);
drivers/tty/serial/sunhv.c +6 -0
@@ -398,6 +398,12 @@ static int sunhv_verify_port(struct uart_port *port, struct serial_struct *ser)static struct uart_port *sunhv_port; +void sunhv_migrate_hvcons_irq (int cpu) +{ +/* Migrate hvcons irq to param cpu */ +irq_force_affinity (sunhv_port->irq,cpumask_of (cpu)); +} +/* Copy 's' into the con_write_page, decoding "\n" into * "\r\n" along the way. We have to return two lengths * because the caller needs to know how much to advance
Signature · 1 line
1.7.1
PATCH v3 3/4 sparc64: Send break twice from console to return to boot prom
In reply to Vijay KumarNow we can also jump to boot prom from sunhv console by sending break twice on console for both running and panicked kernel cases.
--- drivers/tty/serial/sunhv.c | 6 +++++- kernel/panic.c | 3 ++- 2 files changed, 7 insertions(+), 2 deletions(-)
drivers/tty/serial/sunhv.c +5 -1
@@ -116,7 +116,7 @@ static int receive_chars_getchar(struct uart_port *port)static int receive_chars_read (struct uart_port *port) { -int saw_console_brk =0 ; +static int saw_console_brk;int limit =10000 ;while (limit-- >0 ) {
@@ -128,6 +128,9 @@ static int receive_chars_read(struct uart_port *port) bytes_read =0 ;if (stat =CON_BREAK ) { +if (saw_console_brk) +sun_do_break (); +if (uart_handle_break (port))continue ; saw_console_brk =1 ;
@@ -151,6 +154,7 @@ static int receive_chars_read(struct uart_port *port)if (port->sysrq !=0 && *con_read_page) {for (i =0 ; i < bytes_read; i++)uart_handle_sysrq_char (port, con_read_page[i]); + saw_console_brk =0 ; }if (port->state =NULL )
kernel/panic.c +2 -1
@@ -273,7 +273,8 @@ void panic(const char *fmt, ...)extern int stop_a_enabled;/* Make sure the user can actually press Stop-A (L1-A) */ stop_a_enabled =1 ; - pr_emerg("Press Stop-A (L1-A) to return to the boot prom\n"); +pr_emerg ("Press Stop-A (L1-A) from sun keyboard or send break\n" + "twice on console to return to the boot prom\n"); }#endif #if defined(CONFIG_S390 )
Signature · 1 line
1.7.1
PATCH v3 4/4 Documentation/sparc: Steps for sending break on sunhv console
In reply to Vijay KumarDocumented the steps for sending break on sunhv console.
--- Documentation/sparc/console.txt | 9 +++++++++ 1 files changed, 9 insertions(+), 0 deletions(-)
Documentation/sparc/console.txt new file +8 -0
@@ -0,0 +1,9 @@ +Steps for sending 'break' on sunhv console: +=====================+ +On Baremetal: + 1. press Esc + 'B' + +On LDOM: + 1. press Ctrl + ']' + 2. telnet> send break
Signature · 1 line
1.7.1
David Miller
maintainerFrom: Vijay Kumar <[email protected]>
Date: Wed, 1 Feb 2017 11:34:36 -0800
Quoted 1 line
Currently Stop-A (L1A) does not make the kernel switch to OBP on panic.
This is intentional, the kernel prints a message telling the user to press break (L1-A) if they want to drop out of the kernel and we force the break to be allowed by setting stop_a_enabled.
I'm wondering why there is so much effort being directed into BRK behavior.
If you want to break into the OK prompt, have the reboot-cmd environment variable set appropriately, and simply hit BRK and it will work in both ldom and non-ldom environments.
Quoted 7 lines · David Miller
From: Vijay Kumar <[email protected]>
Date: Wed, 1 Feb 2017 11:34:36 -0800Quoted 1 line
Currently Stop-A (L1A) does not make the kernel switch to OBP on panic.
This is intentional, the kernel prints a message telling the user to press break (L1-A) if they want to drop out of the kernel and we force the break to be allowed by setting
stop_a_enabled.
The problem is that pressing BRK after panic does not drop to OK prompt (when stop_a_enabled is set). So the kernel message to press Stop-A to return to boot prom is misleading in this case.
Quoted 2 lines
I'm wondering why there is so much effort being directed into BRK behavior.
User can drop into ok prompt from the running kernel and as well as from the panicked kernel. Pressing single break to jump to ok prompt conflicts with sysrq key combination (from console, BRK + sysrq_key). To be consistent across both the cases, user will have to send BRK twice in order to drop to ok prompt. Does this sound reasonable?
Quoted 4 lines
If you want to break into the OK prompt, have the reboot-cmd environment variable set appropriately, and simply hit BRK and it will work in both ldom and non-ldom environments.
Kernel does not print message "Press Stop-A (L1-A) to ..." for the case when it is expected to reboot on panic. Rather, it goes through different path in panic() when kernel.panic is _not_ set to 0. Here, patch is addressing the case when kernel.panic=0 (i.e not to reboot on panic).
Thanks,
Vijay
David Miller
maintainerFrom: Vijay Kumar <[email protected]>
Date: Wed, 1 Feb 2017 15:20:08 -0600
Quoted 7 lines · David Miller
Quoted 5 lines
From: Vijay Kumar <[email protected]>
Date: Wed, 1 Feb 2017 11:34:36 -0800Quoted 2 lines
Currently Stop-A (L1A) does not make the kernel switch to OBP on panic.
Quoted 8 lines
Quoted 3 lines
This is intentional, the kernel prints a message telling the user to press break (L1-A) if they want to drop out of the kernel and we force the break to be allowed by setting
stop_a_enabled.The problem is that pressing BRK after panic does not drop to OK prompt (when
stop_a_enabledis set). So the kernel message to press Stop-A to return to boot
prom is misleading in this case.
Ok, I ended up applying this patch series, thanks.