From: Paul Durrant <Paul.Durrant@citrix.com>
To: Anthony Perard <anthony.perard@citrix.com>
Cc: "xen-devel@lists.xenproject.org" <xen-devel@lists.xenproject.org>,
Stefano Stabellini <sstabellini@kernel.org>,
"qemu-devel@nongnu.org" <qemu-devel@nongnu.org>,
"qemu-block@nongnu.org" <qemu-block@nongnu.org>
Subject: Re: [PATCH 15/18] xen: add a mechanism to automatically create XenDevice-s...
Date: Thu, 6 Dec 2018 12:36:52 +0000 [thread overview]
Message-ID: <66fe7ee71e9642d3bbbda0f4f900c799@AMSPEX02CL03.citrite.net> (raw)
In-Reply-To: <20181204153510.GW14786@perard.uk.xensource.com>
> -----Original Message-----
> From: Anthony PERARD [mailto:anthony.perard@citrix.com]
> Sent: 04 December 2018 15:35
> To: Paul Durrant <Paul.Durrant@citrix.com>
> Cc: qemu-block@nongnu.org; qemu-devel@nongnu.org; xen-
> devel@lists.xenproject.org; Stefano Stabellini <sstabellini@kernel.org>
> Subject: Re: [PATCH 15/18] xen: add a mechanism to automatically create
> XenDevice-s...
>
> On Wed, Nov 21, 2018 at 03:12:08PM +0000, Paul Durrant wrote:
> > + xen_backend_device_create(BUS(xenbus), type, name, opts,
> &local_err);
> > + qobject_unref(opts);
> > +
> > + if (local_err) {
> > + const char *msg = error_get_pretty(local_err);
> > +
> > + error_report("failed to create '%s' device '%s': %s", type,
> name,
> > + msg);
> > + error_free(local_err);
>
> You can use error_reportf_err() instead of those three calls. I may have
> only suggest error_report_err in a previous patch, but error_reportf_err
> does the error_prepend as well.
>
Ah. I'll go back over the patches and use that where necessary.
> > + }
> > +}
> > +
> > +static void xen_bus_type_enumerate(XenBus *xenbus, const char *type)
> > +{
> > + char *domain_path = g_strdup_printf("backend/%s/%u", type,
> xen_domid);
> > + char **backend;
> > + unsigned int i, n;
> > +
> > + trace_xen_bus_type_enumerate(type);
> > +
> > + backend = xs_directory(xenbus->xsh, XBT_NULL, domain_path, &n);
> > + if (!backend) {
>
> domain_path isn't free here, you probably want a `goto out` which would
> free everything.
Ok.
>
> > + return;
> > + }
> > +
> > @@ -193,6 +302,17 @@ static void xen_bus_realize(BusState *bus, Error
> **errp)
> > notifier_list_init(&xenbus->watch_notifiers);
> > qemu_set_fd_handler(xs_fileno(xenbus->xsh), xen_bus_watch, NULL,
> > xenbus);
> > +
> > + module_call_init(MODULE_INIT_XEN_BACKEND);
> > +
> > + xenbus->backend_watch =
> > + xen_bus_add_watch(xenbus, "", /* domain root node */
> > + "backend", xen_bus_enumerate, xenbus,
> &local_err);
> > + if (local_err) {
> > + error_propagate(errp, local_err);
> > + error_prepend(errp, "failed to set up enumeration watch: ");
>
> You should use error_propagate_prepend instead
> error_propagate;error_prepend. And it looks like there is the same
> mistake in other patches that I haven't notice.
>
Oh, I didn't know about that one either... I've only seen the separate calls used elsewhere.
> Also you probably want goto fail here.
>
Not sure about that. Whilst the bus scan won't happen, it doesn't mean devices can't be added via QMP.
>
> > +static void xen_device_backend_changed(void *opaque)
> > +{
> > + XenDevice *xendev = opaque;
> > + const char *type = object_get_typename(OBJECT(xendev));
> > + enum xenbus_state state;
> > + unsigned int online;
> > +
> > + trace_xen_device_backend_changed(type, xendev->name);
> > +
> > + if (xen_device_backend_scanf(xendev, "state", "%u", &state) != 1) {
> > + state = XenbusStateUnknown;
> > + }
> > +
> > + xen_device_backend_set_state(xendev, state);
>
> It's kind of weird to set the internal state base on the external one
> that something else may have modified. Shouldn't we check that it is
> fine for something else to modify the state and that it is a correct
> transition?
The only thing (apart from this code) that's going to have perms to write the backend state is the toolstack... which is, of course, be definition trusted.
>
> Also aren't we going in a loop by having QEMU set the state, then the
> watch fires again? (Not really a loop since the function _set_state
> check for changes.
No. It's de-bounced inside the set_state function.
>
> Also maybe we should watch for the state changes only when something
> else like libxl creates (ask for) the backend, and ignore changes when
> QEMU did it itself.
I don't think it's necessary to add that complexity.
>
> > +
> > + if (xen_device_backend_scanf(xendev, "online", "%u", &online) != 1)
> {
> > + online = 0;
> > + }
> > +
> > + xen_device_backend_set_online(xendev, !!online);
> > +
> > + /*
> > + * If a backend is still 'online' then its state should be cycled
> > + * back round to InitWait in order for a new frontend instance to
> > + * connect. This may happen when, for example, a frontend driver is
> > + * re-installed or updated.
> > + * If a backend id not 'online' then the device should be
> destroyed.
>
> s/id/is/
Ok.
>
> > + */
> > + if (xendev->backend_online &&
> > + xendev->backend_state == XenbusStateClosed) {
> > + xen_device_backend_set_state(xendev, XenbusStateInitWait);
> > + } else if (!xendev->backend_online &&
> > + (xendev->backend_state == XenbusStateClosed ||
> > + xendev->backend_state == XenbusStateInitialising ||
> > + xendev->backend_state == XenbusStateInitWait ||
> > + xendev->backend_state == XenbusStateUnknown)) {
> > + object_unparent(OBJECT(xendev));
> > + }
> > +}
> > +
> > static void xen_device_backend_create(XenDevice *xendev, Error **errp)
> > {
> > XenBus *xenbus = XEN_BUS(qdev_get_parent_bus(DEVICE(xendev)));
> > @@ -289,12 +463,38 @@ static void xen_device_backend_create(XenDevice
> *xendev, Error **errp)
> > error_propagate(errp, local_err);
> > error_prepend(errp, "failed to create backend: ");
>
> It looks like there is a missing return here.
>
> > }
> > +
> > + xendev->backend_state_watch =
> > + xen_bus_add_watch(xenbus, xendev->backend_path,
> > + "state", xen_device_backend_changed,
> > + xendev, &local_err);
> > + if (local_err) {
> > + error_propagate(errp, local_err);
> > + error_prepend(errp, "failed to watch backend state: ");
>
> You should return here, as local_err mustn't be reused.
>
> > + }
> > +
> > + xendev->backend_online_watch =
> > + xen_bus_add_watch(xenbus, xendev->backend_path,
> > + "online", xen_device_backend_changed,
> > + xendev, &local_err);
> > + if (local_err) {
> > + error_propagate(errp, local_err);
> > + error_prepend(errp, "failed to watch backend online: ");
>
> You probably want a return here, in case there is more code added after.
Yes, there should be returns in all three cases above.
>
> > + }
>
> Other instances of error_propagate;error_prepend to be replaced by
> error_propagate_prepend.
Yes, will do.
Paul
>
> > }
> >
>
> Thanks,
>
> --
> Anthony PERARD
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
next prev parent reply other threads:[~2018-12-06 12:36 UTC|newest]
Thread overview: 78+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <20181121151211.15997-1-paul.durrant@citrix.com>
2018-11-21 15:11 ` [PATCH 01/18] xen: re-name XenDevice to XenLegacyDevice Paul Durrant
2018-11-28 16:06 ` Anthony PERARD
2018-11-21 15:11 ` [PATCH 02/18] xen: introduce new 'XenBus' and 'XenDevice' object hierarchy Paul Durrant
2018-11-21 15:11 ` [PATCH 03/18] xen: introduce 'xen-qdisk' Paul Durrant
2018-11-21 15:11 ` [PATCH 04/18] xen: create xenstore areas for XenDevice-s Paul Durrant
2018-11-21 15:11 ` [PATCH 05/18] xen: add xenstore watcher infratructure Paul Durrant
2018-11-21 15:11 ` [PATCH 06/18] xen: add grant table interface for XenDevice-s Paul Durrant
2018-12-03 15:45 ` Anthony PERARD
2018-12-05 16:12 ` Paul Durrant
2018-11-21 15:12 ` [PATCH 07/18] xen: add event channel " Paul Durrant
2018-11-21 15:12 ` [PATCH 08/18] xen: duplicate xen_disk.c as basis of dataplane/xen-qdisk.c Paul Durrant
2018-12-03 16:35 ` Anthony PERARD
2018-12-03 16:42 ` Anthony PERARD
2018-11-21 15:12 ` [PATCH 09/18] xen: remove unnecessary code from dataplane/xen-qdisk.c Paul Durrant
2018-12-03 16:58 ` Anthony PERARD
2018-11-21 15:12 ` [PATCH 10/18] xen: add header and build dataplane/xen-qdisk.c Paul Durrant
2018-12-03 18:09 ` Anthony PERARD
[not found] ` <20181203180911.GQ14786@perard.uk.xensource.com>
2018-12-05 17:31 ` Paul Durrant
2018-11-21 15:12 ` [PATCH 11/18] xen: remove 'XenBlkDev' and 'blkdev' names from dataplane/xen-qdisk Paul Durrant
2018-11-21 15:12 ` [PATCH 12/18] xen: remove 'ioreq' struct/varable/field names from dataplane/xen-qdisk.c Paul Durrant
2018-12-04 11:34 ` Anthony PERARD
2018-11-21 15:12 ` [PATCH 13/18] xen: purge 'blk' and 'ioreq' from function names in dataplane/xen-qdisk.c Paul Durrant
2018-12-04 12:10 ` Anthony PERARD
2018-12-05 17:28 ` Paul Durrant
2018-11-21 15:12 ` [PATCH 14/18] xen: add implementations of xen-qdisk connect and disconnect functions Paul Durrant
2018-11-21 15:12 ` [PATCH 15/18] xen: add a mechanism to automatically create XenDevice-s Paul Durrant
2018-12-04 15:35 ` Anthony PERARD
2018-12-06 12:36 ` Paul Durrant [this message]
2018-12-06 15:24 ` Anthony PERARD
[not found] ` <20181206152406.GC18875@perard.uk.xensource.com>
2018-12-06 15:36 ` Paul Durrant
2018-11-21 15:12 ` [PATCH 16/18] xen: automatically create XenQdiskDevice-s Paul Durrant
2018-11-21 15:12 ` [PATCH 17/18] MAINTAINERS: add myself as a Xen maintainer Paul Durrant
2018-11-27 19:05 ` Stefano Stabellini
2018-11-29 14:00 ` [Qemu-devel] " Philippe Mathieu-Daudé
[not found] ` <a247dae0-a766-7f97-c16d-07a32572f35d@redhat.com>
2018-11-29 14:01 ` Paul Durrant
2018-12-04 16:42 ` Anthony PERARD
2018-11-21 15:12 ` [PATCH 18/18] xen: remove the legacy 'xen_disk' backend Paul Durrant
[not found] ` <20181121151211.15997-3-paul.durrant@citrix.com>
2018-11-28 16:19 ` [Qemu-block] [PATCH 02/18] xen: introduce new 'XenBus' and 'XenDevice' object hierarchy Kevin Wolf
[not found] ` <20181128161917.GE4222@dhcp-200-186.str.redhat.com>
2018-11-28 16:26 ` Paul Durrant
2018-11-28 16:28 ` Paul Durrant
2018-11-28 16:28 ` Stefano Stabellini
2018-11-28 16:29 ` Paul Durrant
2018-11-28 16:39 ` Kevin Wolf
2018-11-28 16:45 ` Paul Durrant
[not found] ` <5327ba765089439caf8119de49c3a399@AMSPEX02CL03.citrite.net>
2018-11-28 16:46 ` Paul Durrant
2018-11-29 9:04 ` Kevin Wolf
2018-11-28 17:01 ` [Qemu-devel] " Eric Blake
2018-11-28 17:04 ` Paul Durrant
2018-11-28 17:10 ` Anthony PERARD
2018-11-28 17:17 ` Paul Durrant
2018-11-28 17:32 ` Anthony PERARD
[not found] ` <20181121151211.15997-4-paul.durrant@citrix.com>
2018-11-29 16:05 ` [PATCH 03/18] xen: introduce 'xen-qdisk' Anthony PERARD
[not found] ` <20181129160541.GG14786@perard.uk.xensource.com>
2018-12-04 15:20 ` Paul Durrant
[not found] ` <311937f3c787461686f2ebf68a70a326@AMSPEX02CL03.citrite.net>
2018-12-04 15:49 ` Anthony PERARD
[not found] ` <20181204154908.GX14786@perard.uk.xensource.com>
2018-12-04 15:50 ` Paul Durrant
2018-12-04 17:14 ` Paul Durrant
[not found] ` <20181121151211.15997-5-paul.durrant@citrix.com>
2018-11-29 18:48 ` [PATCH 04/18] xen: create xenstore areas for XenDevice-s Anthony PERARD
[not found] ` <20181129184841.GJ14786@perard.uk.xensource.com>
2018-12-05 12:05 ` Paul Durrant
[not found] ` <589a4488f1ba4e4496775cc06bda291d@AMSPEX02CL03.citrite.net>
2018-12-05 12:43 ` Paul Durrant
[not found] ` <d5bf17f689a34e69b9e3b7b92bcfe4fd@AMSPEX02CL03.citrite.net>
2018-12-05 13:58 ` Anthony PERARD
[not found] ` <20181205135839.GB1259@perard.uk.xensource.com>
2018-12-05 14:24 ` Paul Durrant
2018-12-05 16:28 ` Anthony PERARD
[not found] ` <20181121151211.15997-6-paul.durrant@citrix.com>
2018-12-03 14:42 ` [PATCH 05/18] xen: add xenstore watcher infratructure Anthony PERARD
[not found] ` <20181203144231.GK14786@perard.uk.xensource.com>
2018-12-05 15:24 ` Paul Durrant
[not found] ` <20181121151211.15997-8-paul.durrant@citrix.com>
2018-12-03 16:24 ` [PATCH 07/18] xen: add event channel interface for XenDevice-s Anthony PERARD
2018-12-04 14:24 ` Anthony PERARD
2018-12-05 16:16 ` Paul Durrant
[not found] ` <20181121151211.15997-12-paul.durrant@citrix.com>
2018-12-04 11:05 ` [PATCH 11/18] xen: remove 'XenBlkDev' and 'blkdev' names from dataplane/xen-qdisk Anthony PERARD
[not found] ` <20181121151211.15997-15-paul.durrant@citrix.com>
2018-11-28 16:34 ` [PATCH 14/18] xen: add implementations of xen-qdisk connect and disconnect functions Kevin Wolf
2018-11-28 16:40 ` Paul Durrant
2018-11-29 9:00 ` Kevin Wolf
[not found] ` <20181129090046.GA4797@linux.fritz.box>
2018-11-29 9:33 ` Paul Durrant
[not found] ` <bdee6326dd3041f58202220a50abf9c8@AMSPEX02CL03.citrite.net>
2018-11-29 10:46 ` Kevin Wolf
2018-11-29 10:47 ` Paul Durrant
2018-12-04 12:33 ` Anthony PERARD
[not found] ` <20181204123349.GU14786@perard.uk.xensource.com>
2018-12-06 12:27 ` Paul Durrant
[not found] ` <20181121151211.15997-17-paul.durrant@citrix.com>
2018-12-04 16:40 ` [PATCH 16/18] xen: automatically create XenQdiskDevice-s Anthony PERARD
2018-12-06 13:06 ` Paul Durrant
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=66fe7ee71e9642d3bbbda0f4f900c799@AMSPEX02CL03.citrite.net \
--to=paul.durrant@citrix.com \
--cc=anthony.perard@citrix.com \
--cc=qemu-block@nongnu.org \
--cc=qemu-devel@nongnu.org \
--cc=sstabellini@kernel.org \
--cc=xen-devel@lists.xenproject.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).