[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Thu, 12 Dec 2019 13:37:43 +0000
From: Tianyu Lan <Tianyu.Lan@...rosoft.com>
To: vkuznets <vkuznets@...hat.com>,
"lantianyu1986@...il.com" <lantianyu1986@...il.com>,
KY Srinivasan <kys@...rosoft.com>,
Haiyang Zhang <haiyangz@...rosoft.com>,
Stephen Hemminger <sthemmin@...rosoft.com>,
"sashal@...nel.org" <sashal@...nel.org>,
Michael Kelley <mikelley@...rosoft.com>
CC: "linux-hyperv@...r.kernel.org" <linux-hyperv@...r.kernel.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
"eric.devolder@...cle.com" <eric.devolder@...cle.com>
Subject: RE: [EXTERNAL] Re: [RFC PATCH 4/4] x86/Hyper-V: Add memory hot remove
function
> From: Vitaly Kuznetsov <vkuznets@...hat.com>
>
> > From: Tianyu Lan <Tianyu.Lan@...rosoft.com>
> >
> > @@ -376,6 +391,27 @@ struct dm_hot_add_response {
> > __u32 result;
> > } __packed;
> >
> > +struct dm_hot_remove {
> > + struct dm_header hdr;
> > + __u32 virtual_node;
> > + __u32 page_count;
> > + __u32 qos_flags;
> > + __u32 reservedZ;
> > +} __packed;
> > +
> > +struct dm_hot_remove_response {
> > + struct dm_header hdr;
> > + __u32 result;
> > + __u32 range_count;
> > + __u64 more_pages:1;
> > + __u64 reservedz:63;
> > + union dm_mem_page_range range_array[]; } __packed;
> > +
> > +#define DM_REMOVE_QOS_LARGE (1 << 0)
> > +#define DM_REMOVE_QOS_LOCAL (1 << 1)
> > +#define DM_REMOVE_QoS_MASK (0x3)
>
> Capitalize 'QoS' to make it match previous two lines please.
>
Yes, Will fix it.
> > +
> > /*
> > * Types of information sent from host to the guest.
> > */
> > @@ -457,6 +493,13 @@ struct hot_add_wrk {
> > struct work_struct wrk;
> > };
> >
> > +struct hot_remove_wrk {
> > + __u32 virtual_node;
> > + __u32 page_count;
> > + __u32 qos_flags;
> > + struct work_struct wrk;
> > +};
> > +
> > static bool hot_add = true;
> > static bool do_hot_add;
> > /*
> > @@ -489,6 +532,7 @@ enum hv_dm_state {
> > DM_BALLOON_UP,
> > DM_BALLOON_DOWN,
> > DM_HOT_ADD,
> > + DM_HOT_REMOVE,
> > DM_INIT_ERROR
> > };
> >
> > @@ -515,11 +559,13 @@ struct hv_dynmem_device {
> > * State to manage the ballooning (up) operation.
> > */
> > struct balloon_state balloon_wrk;
> > + struct balloon_state unballoon_wrk;
> >
> > /*
> > * State to execute the "hot-add" operation.
>
> This comment is stale now.
>
Will update. Thanks.
> > */
> > struct hot_add_wrk ha_wrk;
> > + struct hot_remove_wrk hr_wrk;
>
> Do we actually want to work struct and all the problems with their
> serialization? Can we get away with one?
I think it's possible to just use one work to handle all kind of msgs
with a work struct which contains parameters for all dm msgs and identify
the msg type in the work callback function.
Powered by blists - more mailing lists