This is an automated email from Gerrit. Tim Newsome ([email protected]) just uploaded a new patch set to Gerrit, which you can find at http://openocd.zylin.com/6362
-- gerrit commit f63a4b1b790058aa494389df8a3baebae30c38ab Author: Tim Newsome <[email protected]> Date: Thu Jul 8 17:43:00 2021 -0700 Combine register lists of smp targets. This is helpful when you want to pretend to gdb that your heterogeneous multicore system is homogeneous, because gdb cannot handle heterogeneous systems. This won't always works, but works fine if e.g. one of the cores has an FPU while the other does not. (Specifically, HiFive Unleashed has 1 core with no FPU, plus 4 cores with an FPU.) Signed-off-by: Tim Newsome <[email protected]> Change-Id: I05ff4c28646778fbc00327bc510be064bfe6c9f0 diff --git a/src/server/gdb_server.c b/src/server/gdb_server.c index 9ac982f..df4a987 100644 --- a/src/server/gdb_server.c +++ b/src/server/gdb_server.c @@ -2251,6 +2251,78 @@ static int get_reg_features_list(struct target *target, char const **feature_lis return ERROR_OK; } +/* Create a register list that's the union of all the registers of the SMP + * group this target is in. If the target is not part of an SMP group, this + * returns the same as target_get_gdb_reg_list_noread(). + */ +static int smp_reg_list_noread(struct target *target, + struct reg **combined_list[], int *combined_list_size, + enum target_register_class reg_class) +{ + if (!target->smp) + return target_get_gdb_reg_list_noread(target, combined_list, + combined_list_size, REG_CLASS_ALL); + + int combined_allocated = 256; + *combined_list = malloc(combined_allocated * sizeof(struct reg *)); + if (*combined_list == NULL) { + LOG_ERROR("malloc(%d) failed", (int) (combined_allocated * sizeof(struct reg *))); + return ERROR_FAIL; + } + *combined_list_size = 0; + struct target_list *head; + foreach_smp_target(head, target->head) { + struct reg **reg_list = NULL; + int reg_list_size; + int result = target_get_gdb_reg_list_noread(head->target, ®_list, + ®_list_size, reg_class); + if (result != ERROR_OK) { + free(*combined_list); + return result; + } + for (int i = 0; i < reg_list_size; i++) { + bool found = false; + struct reg *a = reg_list[i]; + if (a->exist) { + /* Nested loop makes this O(n^2), but this entire function with + * 5 RISC-V targets takes just 2ms on my computer. Fast enough + * for me. */ + for (int j = 0; j < *combined_list_size; j++) { + struct reg *b = (*combined_list)[j]; + if (!strcmp(a->name, b->name)) { + found = true; + if (a->size != b->size) { + LOG_ERROR("SMP register %s is %d bits on one " + "target, but %d bits on another target.", + a->name, a->size, b->size); + free(reg_list); + free(*combined_list); + return ERROR_FAIL; + } + break; + } + } + if (!found) { + LOG_DEBUG("[%s] %s not found in combined list", target_name(target), a->name); + if (*combined_list_size >= combined_allocated) { + combined_allocated *= 2; + *combined_list = realloc(*combined_list, combined_allocated * sizeof(struct reg *)); + if (*combined_list == NULL) { + LOG_ERROR("realloc(%d) failed", (int) (combined_allocated * sizeof(struct reg *))); + return ERROR_FAIL; + } + } + (*combined_list)[*combined_list_size] = a; + (*combined_list_size)++; + } + } + } + free(reg_list); + } + + return ERROR_OK; +} + static int gdb_generate_target_description(struct target *target, char **tdesc_out) { int retval = ERROR_OK; @@ -2264,8 +2336,8 @@ static int gdb_generate_target_description(struct target *target, char **tdesc_o int size = 0; - retval = target_get_gdb_reg_list_noread(target, ®_list, - ®_list_size, REG_CLASS_ALL); + retval = smp_reg_list_noread(target, ®_list, ®_list_size, + REG_CLASS_ALL); if (retval != ERROR_OK) { LOG_ERROR("get register list failed"); --
