[Kgdb-bugreport] [PATCH] kgdb_roundup_cpus() with 'O' packets
Status: Beta
Brought to you by:
jwessel
Menu
▾
▴
[Kgdb-bugreport] [PATCH] kgdb_roundup_cpus() with 'O' packets
|
From: Tom R. <tr...@ke...> - 2004-11-30 16:17:14
|
So, here's what I'd like to commit. We always call kgdb_roundup_cpus(),
we provide a weak version, and a send_IPI_allbutself(...) version on
i386/x86_64, which can be expanded to be more robust if needed, wait 'a
bit' to try and catch NR_CPUS and if we fail, we send an 'O' packet, and
only an 'O' packet, once. This doesn't get lost by GDB since we've
waited for GDB to make a request, and the protocol says that on a '?'
GDB must parse O packets and wait for a real answer to come, so it's
protocol-safe. I've changed my mind about printk'ing, since it's too
much of a headache to make sure we don't accidently goof the machine,
and printk isn't 'reliable' here.
--- linux-2.6.10-rc2/include/linux/kgdb.h
+++ linux-2.6.10-rc2/include/linux/kgdb.h
@@ -115,7 +115,7 @@
* kgdb hook.
*/
extern int kgdb_arch_handle_exception(int vector, int signo, int err_code,
- char *InBuffer, char *outBuffer,
+ char *inBuffer, char *outBuffer,
struct pt_regs *regs);
/* Optional functions. */
@@ -123,6 +123,7 @@
extern void kgdb_disable_hw_debug(struct pt_regs *regs);
extern void kgdb_post_master_code(struct pt_regs *regs, int eVector,
int err_code);
+extern void kgdb_roundup_cpus(void);
extern int kgdb_set_hw_break(unsigned long addr);
extern int kgdb_remove_hw_break(unsigned long addr);
extern void kgdb_remove_all_hw_break(void);
--- linux-2.6.10-rc2/kernel/kgdb.c
+++ linux-2.6.10-rc2/kernel/kgdb.c
@@ -59,6 +59,8 @@
extern int pid_max;
extern int pidhash_init_done;
+/* How many times to count all of the waiting CPUs */
+#define ROUNDUP_WAIT 64000
#define BUF_THREAD_ID_SIZE 16
/*
@@ -195,6 +197,18 @@
}
/**
+ * kgdb_roundup_cpus - Get other CPUs into a holding pattern
+ *
+ * On SMP systems, we need to get the attention of the other CPUs
+ * and get them be in a known state. This should do what is needed
+ * to get the other CPUs to call kgdb_wait().
+ */
+void __attribute__ ((weak))
+ kgdb_roundup_cpus(void)
+{
+}
+
+/**
* kgdb_shadowinfo - Get shadowed information on @threadid.
* @regs: The &struct pt_regs of the current process.
* @buffer: A buffer of %BUFMAX size.
@@ -543,7 +557,7 @@
}
/* Wait till master processor is done with debugging */
- spin_lock(slavecpulocks + processor);
+ spin_lock(slavecpulocks[processor]);
/* This has been taken from x86 kgdb implementation and
* will be needed by architectures that have SMP support
@@ -555,7 +569,7 @@
/* Signal the master processor that we are done */
procindebug[processor] = 0;
- spin_unlock(slavecpulocks + processor);
+ spin_unlock(slavecpulocks[processor]);
local_irq_restore(flags);
}
#endif
@@ -697,6 +711,42 @@
return pid_max + smp_processor_id();
}
+#define MAXOUT ((BUFMAX - 2) / 2)
+
+static char gdbmsgbuf[BUFMAX + 1];
+static void kgdb_msg_write(const char *s)
+{
+ int i;
+ int wcount, len;
+ char *bufptr;
+
+ /* 'O'utput */
+ gdbmsgbuf[0] = 'O';
+
+ /* How long of a string. */
+ len = strlen(s);
+
+ /* Fill and send buffers... */
+ while (len > 0) {
+ bufptr = gdbmsgbuf + 1;
+
+ /* Calculate how many this time */
+ wcount = (len > MAXOUT) ? MAXOUT : len;
+
+ /* Pack in hex chars */
+ for (i = 0; i < wcount; i++)
+ bufptr = pack_hex_byte(bufptr, s[i]);
+ *bufptr = '\0';
+
+ /* Move up */
+ s += wcount;
+ len -= wcount;
+
+ /* Write packet */
+ put_packet(gdbmsgbuf);
+ }
+}
+
/*
* This function does all command procesing for interfacing to gdb.
*
@@ -704,6 +754,10 @@
* interface locks, if any (begin_session)
* kgdb lock (debugger_active)
*
+ * Note that since we can be in here prior to our cpumask being filled
+ * out, which err on the side of caution and loop over NR_CPUS instead
+ * of a for_each_online_cpu.
+ *
*/
int kgdb_handle_exception(int exVector, int signo, int err_code,
struct pt_regs *linux_regs)
@@ -718,7 +772,7 @@
unsigned procid;
int numshadowth = num_online_cpus() + kgdb_ops->shadowth;
long kgdb_usethreadid = 0;
- int error = 0;
+ int error = 0, all_cpus_rounded = 0;
struct pt_regs *shadowregs;
int processor = smp_processor_id();
void * local_debuggerinfo;
@@ -753,20 +807,37 @@
goto acquirelock;
}
- kgdb_info[processor].debuggerinfo = linux_regs;
+ kgdb_info[processor].debuggerinfo = linux_regs;
kgdb_info[processor].task = current;
kgdb_disable_hw_debug(linux_regs);
- if(!debugger_step || !kgdb_contthread) {
- for (i = 0; i < num_online_cpus(); i++)
+ if (!debugger_step || !kgdb_contthread)
+ for (i = 0; i < NR_CPUS; i++)
spin_lock(&slavecpulocks[i]);
- }
+
+ /* Make sure we get the other CPUs */
+ kgdb_roundup_cpus();
/* spin_lock code is good enough as a barrier so we don't
* need one here */
procindebug[smp_processor_id()] = 1;
+ /* Wait a reasonable time for the other CPUs to be notified and
+ * be waiting for us. Very early on this could be imperfect
+ * as num_online_cpus() could be 0.*/
+ for (i = 0; i < ROUNDUP_WAIT; i++) {
+ int cpu, num = 0;
+ for (cpu = 0; cpu < NR_CPUS; cpu++) {
+ if (procindebug[cpu])
+ num++;
+ }
+ if (num >= num_online_cpus()) {
+ all_cpus_rounded = 1;
+ break;
+ }
+ }
+
/* Clear the out buffer. */
memset(remcom_out_buffer, 0, sizeof(remcom_out_buffer));
@@ -809,6 +880,12 @@
* we clear out our breakpoints now incase
* GDB is reconnecting. */
remove_all_break();
+ /* Also, if we haven't been able to roundup all
+ * CPUs, send an 'O' packet informing the user
+ * as much. Only need to do this once. */
+ if (!all_cpus_rounded)
+ kgdb_msg_write("Not all CPUs have been"
+ "synced for KGDB\n");
remcom_out_buffer[0] = 'S';
remcom_out_buffer[1] = hexchars[signo >> 4];
remcom_out_buffer[2] = hexchars[signo % 16];
@@ -915,12 +992,14 @@
kgdb_connected = 0;
}
put_packet(remcom_out_buffer);
+ all_cpus_rounded = 0;
goto default_handle;
case 'k':
/* Don't care about error from remove_all_break */
remove_all_break();
kgdb_connected = 0;
+ all_cpus_rounded = 0;
goto default_handle;
/* query */
@@ -1144,12 +1223,12 @@
procindebug[smp_processor_id()] = 0;
if (!debugger_step || !kgdb_contthread) {
- for (i = 0; i < num_online_cpus(); i++)
- spin_unlock(&slavecpulocks[i]);
+ for (i = 0; i < NR_CPUS; i++)
+ spin_unlock(&slavecpulocks[i]);
/* Wait till all the processors have quit
- * from the debugger
- */
- for_each_online_cpu (i) {
+ * from the debugger.
+ */
+ for (i = 0; i < NR_CPUS; i++) {
while (procindebug[i]) {
int j = 10; /* an arbitrary number */
--- linux-2.6.10-rc2/arch/i386/kernel/kgdb.c
+++ linux-2.6.10-rc2/arch/i386/kernel/kgdb.c
@@ -37,9 +37,12 @@
#include <asm/ptrace.h> /* for linux pt_regs struct */
#include <linux/kgdb.h>
#include <linux/init.h>
+#include <asm/apicdef.h>
#include <asm/desc.h>
#include <asm/kdebug.h>
+#include "mach_ipi.h"
+
/* Put the error code here just in case the user cares. */
int gdb_i386errcode;
/* Likewise, the vector number here (since GDB only gets the signal
@@ -194,6 +197,11 @@
gdb_i386errcode = err_code;
}
+void kgdb_roundup_cpus(void)
+{
+ send_IPI_allbutself(APIC_DM_NMI);
+}
+
int kgdb_arch_handle_exception(int exceptionVector, int signo,
int err_code, char *remcom_in_buffer,
char *remcom_out_buffer,
--- linux-2.6.10-rc2/arch/x86_64/kernel/kgdb.c
+++ linux-2.6.10-rc2/arch/x86_64/kernel/kgdb.c
@@ -39,6 +39,7 @@
#include <asm/ptrace.h> /* for linux pt_regs struct */
#include <linux/kgdb.h>
#include <linux/init.h>
+#include <asm/mach_apic.h>
#include <asm/kdebug.h>
/* Put the error code here just in case the user cares. */
@@ -260,6 +261,11 @@
gdb_x86_64errcode = err_code;
}
+void kgdb_roundup_cpus(void)
+{
+ send_IPI_allbutself(APIC_DM_NMI);
+}
+
int kgdb_arch_handle_exception(int exceptionVector, int signo, int err_code,
char *remcomInBuffer, char *remcomOutBuffer,
struct pt_regs *linux_regs)
--
Tom Rini
http://gate.crashing.org/~trini/
|