[Crash-utility] [PATCH 3/3] xen: adjust to new scheduler structures
Juergen Gross
jgross at suse.com
Wed Mar 15 05:57:42 UTC 2023
On 15.03.23 06:54, lijiang wrote:
> On Mon, Mar 13, 2023 at 9:07 PM <crash-utility-request at redhat.com
> <mailto:crash-utility-request at redhat.com>> wrote:
>
> Send Crash-utility mailing list submissions to
> Date: Mon, 13 Mar 2023 14:01:12 +0100
> From: Juergen Gross <jgross at suse.com <mailto:jgross at suse.com>>
> To: crash-utility at redhat.com <mailto:crash-utility at redhat.com>
> Subject: [Crash-utility] [PATCH 3/3] xen: adjust to new scheduler
> structures
> Message-ID: <20230313130112.15353-4-jgross at suse.com
> <mailto:20230313130112.15353-4-jgross at suse.com>>
> Content-Type: text/plain; charset="US-ASCII"; x-default=true
>
> There has been a significant modification regarding scheduler data in
> the Xen hypervisor. Adapt to new structures and removed fields.
>
> I would suggest adding the related hypervisor commit here.
>
> Note that this is only the bare minimum to not let crash error out when
> opening a vmcore in Xen mode with a recent Xen version.
>
> Signed-off-by: Juergen Gross <jgross at suse.com <mailto:jgross at suse.com>>
> ---
> xen_hyper.c | 67 +++++++++++++++++++++++++++++++++---------------
> xen_hyper_defs.h | 4 ++-
> 2 files changed, 49 insertions(+), 22 deletions(-)
>
> diff --git a/xen_hyper.c b/xen_hyper.c
> index 72720e2..4c884dd 100644
> --- a/xen_hyper.c
> +++ b/xen_hyper.c
> @@ -417,13 +417,21 @@ void
> xen_hyper_misc_init(void)
> {
> XEN_HYPER_STRUCT_SIZE_INIT(schedule_data, "schedule_data");
> - XEN_HYPER_MEMBER_OFFSET_INIT(schedule_data_schedule_lock,
> "schedule_data", "schedule_lock");
> - XEN_HYPER_MEMBER_OFFSET_INIT(schedule_data_curr, "schedule_data",
> "curr");
> - if (MEMBER_EXISTS("schedule_data", "idle"))
> - XEN_HYPER_MEMBER_OFFSET_INIT(schedule_data_idle,
> "schedule_data", "idle");
> - XEN_HYPER_MEMBER_OFFSET_INIT(schedule_data_sched_priv,
> "schedule_data", "sched_priv");
> - XEN_HYPER_MEMBER_OFFSET_INIT(schedule_data_s_timer, "schedule_data",
> "s_timer");
> - XEN_HYPER_MEMBER_OFFSET_INIT(schedule_data_tick, "schedule_data",
> "tick");
> + XEN_HYPER_STRUCT_SIZE_INIT(sched_resource, "sched_resource");
> + if (XEN_HYPER_VALID_SIZE(schedule_data)) {
> + XEN_HYPER_MEMBER_OFFSET_INIT(schedule_data_schedule_lock,
> "schedule_data", "schedule_lock");
> + XEN_HYPER_MEMBER_OFFSET_INIT(schedule_data_curr,
> "schedule_data", "curr");
> + if (MEMBER_EXISTS("schedule_data", "idle"))
> + XEN_HYPER_MEMBER_OFFSET_INIT(schedule_data_idle,
> "schedule_data", "idle");
> + XEN_HYPER_MEMBER_OFFSET_INIT(schedule_data_sched_priv,
> "schedule_data", "sched_priv");
> + XEN_HYPER_MEMBER_OFFSET_INIT(schedule_data_s_timer,
> "schedule_data", "s_timer");
> + XEN_HYPER_MEMBER_OFFSET_INIT(schedule_data_tick,
> "schedule_data", "tick");
> + } else if (XEN_HYPER_VALID_SIZE(sched_resource)) {
> + XEN_HYPER_MEMBER_OFFSET_INIT(schedule_data_schedule_lock,
> "sched_resource", "schedule_lock");
> + XEN_HYPER_MEMBER_OFFSET_INIT(schedule_data_curr,
> "sched_resource", "curr");
> + XEN_HYPER_MEMBER_OFFSET_INIT(schedule_data_sched_priv,
> "sched_resource", "sched_priv");
> + XEN_HYPER_MEMBER_OFFSET_INIT(schedule_data_s_timer,
> "sched_resource", "s_timer");
> + }
>
> XEN_HYPER_STRUCT_SIZE_INIT(scheduler, "scheduler");
> XEN_HYPER_MEMBER_OFFSET_INIT(scheduler_name, "scheduler", "name");
> @@ -467,6 +475,7 @@ xen_hyper_schedule_init(void)
> long *schedulers_buf;
> int nr_schedulers;
> struct xen_hyper_sched_context *schc;
> + long buf_size;
> char *buf;
> char opt_name_buf[XEN_HYPER_OPT_SCHED_SIZE];
> int i, cpuid, flag;
> @@ -561,28 +570,43 @@ xen_hyper_schedule_init(void)
> }
> BZERO(xhscht->sched_context_array,
> sizeof(struct xen_hyper_sched_context) * XEN_HYPER_MAX_CPUS());
> - buf = GETBUF(XEN_HYPER_SIZE(schedule_data));
> - if (symbol_exists("per_cpu__schedule_data")) {
> + if (symbol_exists("per_cpu__sched_res")) {
> + addr = symbol_value("per_cpu__sched_res");
> + buf_size = XEN_HYPER_SIZE(sched_resource);
> + flag = 0;
> + } else if (symbol_exists("per_cpu__schedule_data")) {
> addr = symbol_value("per_cpu__schedule_data");
> - flag = TRUE;
> + buf_size = XEN_HYPER_SIZE(schedule_data);
> + flag = 1;
> } else {
> addr = symbol_value("schedule_data");
> - flag = FALSE;
> + buf_size = XEN_HYPER_SIZE(schedule_data);
> + flag = 2;
> }
> + buf = GETBUF(buf_size);
> for_cpu_indexes(i, cpuid)
> {
> schc = &xhscht->sched_context_array[cpuid];
> if (flag) {
> - schc->schedule_data =
> - xen_hyper_per_cpu(addr, i);
> + if (flag == 1) {
> + schc->schedule_data =
> + xen_hyper_per_cpu(addr, i);
> + } else {
> + schc->schedule_data = addr +
> + XEN_HYPER_SIZE(schedule_data) * i;
> + }
> + if (!readmem(schc->schedule_data,
> + KVADDR, buf, XEN_HYPER_SIZE(schedule_data),
> + "schedule_data", RETURN_ON_ERROR)) {
> + error(FATAL, "cannot read schedule_data.\n");
> + }
>
>
> As we mentioned in patch 2/3, the readmem(..., FAULT_ON_ERROR) looks better for
> this case.
Okay.
Juergen
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_0xB0DE9DD628BF132F.asc
Type: application/pgp-keys
Size: 3098 bytes
Desc: OpenPGP public key
URL: <http://listman.redhat.com/archives/crash-utility/attachments/20230315/0a893657/attachment-0001.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature
Type: application/pgp-signature
Size: 495 bytes
Desc: OpenPGP digital signature
URL: <http://listman.redhat.com/archives/crash-utility/attachments/20230315/0a893657/attachment-0001.sig>
More information about the Crash-utility
mailing list