2020-01-16 14:17:58

by Tianyu Lan

[permalink] [raw]
Subject: [PATCH V2] x86/Hyper-V: Balloon up according to request page number

From: Tianyu Lan <[email protected]>

Current code has assumption that balloon request memory size aligns
with 2MB. But actually Hyper-V doesn't guarantee such alignment. When
balloon driver receives non-aligned balloon request, it produces warning
and balloon up more memory than requested in order to keep 2MB alignment.
Remove the warning and balloon up memory according to actual requested
memory size.

Fixes: f6712238471a ("hv: hv_balloon: avoid memory leak on alloc_error of 2MB memory block")
Cc: [email protected]
Signed-off-by: Tianyu Lan <[email protected]>
---
Change since v2:
- Change logic of switching alloc_unit from 2MB to 4KB
in the balloon_up() to avoid redundant iteration when
handle non-aligned page request.
- Remove 2MB alignment operation and comment in balloon_up()
---
drivers/hv/hv_balloon.c | 12 ++++--------
1 file changed, 4 insertions(+), 8 deletions(-)

diff --git a/drivers/hv/hv_balloon.c b/drivers/hv/hv_balloon.c
index 7f3e7ab22d5d..536807efbc35 100644
--- a/drivers/hv/hv_balloon.c
+++ b/drivers/hv/hv_balloon.c
@@ -1684,7 +1684,7 @@ static unsigned int alloc_balloon_pages(struct hv_dynmem_device *dm,
if (num_pages < alloc_unit)
return 0;

- for (i = 0; (i * alloc_unit) < num_pages; i++) {
+ for (i = 0; i < num_pages / alloc_unit; i++) {
if (bl_resp->hdr.size + sizeof(union dm_mem_page_range) >
HV_HYP_PAGE_SIZE)
return i * alloc_unit;
@@ -1722,7 +1722,7 @@ static unsigned int alloc_balloon_pages(struct hv_dynmem_device *dm,

}

- return num_pages;
+ return i * alloc_unit;
}

static void balloon_up(union dm_msg_info *msg_info)
@@ -1737,9 +1737,6 @@ static void balloon_up(union dm_msg_info *msg_info)
long avail_pages;
unsigned long floor;

- /* The host balloons pages in 2M granularity. */
- WARN_ON_ONCE(num_pages % PAGES_IN_2M != 0);
-
/*
* We will attempt 2M allocations. However, if we fail to
* allocate 2M chunks, we will go back to PAGE_SIZE allocations.
@@ -1749,14 +1746,13 @@ static void balloon_up(union dm_msg_info *msg_info)
avail_pages = si_mem_available();
floor = compute_balloon_floor();

- /* Refuse to balloon below the floor, keep the 2M granularity. */
+ /* Refuse to balloon below the floor. */
if (avail_pages < num_pages || avail_pages - num_pages < floor) {
pr_warn("Balloon request will be partially fulfilled. %s\n",
avail_pages < num_pages ? "Not enough memory." :
"Balloon floor reached.");

num_pages = avail_pages > floor ? (avail_pages - floor) : 0;
- num_pages -= num_pages % PAGES_IN_2M;
}

while (!done) {
@@ -1770,7 +1766,7 @@ static void balloon_up(union dm_msg_info *msg_info)
num_ballooned = alloc_balloon_pages(&dm_device, num_pages,
bl_resp, alloc_unit);

- if (alloc_unit != 1 && num_ballooned == 0) {
+ if (alloc_unit != 1 && num_ballooned != num_pages) {
alloc_unit = 1;
continue;
}
--
2.14.5


2020-01-16 16:27:49

by Vitaly Kuznetsov

[permalink] [raw]
Subject: Re: [PATCH V2] x86/Hyper-V: Balloon up according to request page number

[email protected] writes:

> From: Tianyu Lan <[email protected]>
>
> Current code has assumption that balloon request memory size aligns
> with 2MB. But actually Hyper-V doesn't guarantee such alignment. When
> balloon driver receives non-aligned balloon request, it produces warning
> and balloon up more memory than requested in order to keep 2MB alignment.
> Remove the warning and balloon up memory according to actual requested
> memory size.
>
> Fixes: f6712238471a ("hv: hv_balloon: avoid memory leak on alloc_error of 2MB memory block")
> Cc: [email protected]
> Signed-off-by: Tianyu Lan <[email protected]>
> ---
> Change since v2:
> - Change logic of switching alloc_unit from 2MB to 4KB
> in the balloon_up() to avoid redundant iteration when
> handle non-aligned page request.
> - Remove 2MB alignment operation and comment in balloon_up()
> ---
> drivers/hv/hv_balloon.c | 12 ++++--------
> 1 file changed, 4 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/hv/hv_balloon.c b/drivers/hv/hv_balloon.c
> index 7f3e7ab22d5d..536807efbc35 100644
> --- a/drivers/hv/hv_balloon.c
> +++ b/drivers/hv/hv_balloon.c
> @@ -1684,7 +1684,7 @@ static unsigned int alloc_balloon_pages(struct hv_dynmem_device *dm,
> if (num_pages < alloc_unit)
> return 0;
>
> - for (i = 0; (i * alloc_unit) < num_pages; i++) {
> + for (i = 0; i < num_pages / alloc_unit; i++) {
> if (bl_resp->hdr.size + sizeof(union dm_mem_page_range) >
> HV_HYP_PAGE_SIZE)
> return i * alloc_unit;
> @@ -1722,7 +1722,7 @@ static unsigned int alloc_balloon_pages(struct hv_dynmem_device *dm,
>
> }
>
> - return num_pages;
> + return i * alloc_unit;
> }
>
> static void balloon_up(union dm_msg_info *msg_info)
> @@ -1737,9 +1737,6 @@ static void balloon_up(union dm_msg_info *msg_info)
> long avail_pages;
> unsigned long floor;
>
> - /* The host balloons pages in 2M granularity. */
> - WARN_ON_ONCE(num_pages % PAGES_IN_2M != 0);
> -
> /*
> * We will attempt 2M allocations. However, if we fail to
> * allocate 2M chunks, we will go back to PAGE_SIZE allocations.
> @@ -1749,14 +1746,13 @@ static void balloon_up(union dm_msg_info *msg_info)
> avail_pages = si_mem_available();
> floor = compute_balloon_floor();
>
> - /* Refuse to balloon below the floor, keep the 2M granularity. */
> + /* Refuse to balloon below the floor. */
> if (avail_pages < num_pages || avail_pages - num_pages < floor) {
> pr_warn("Balloon request will be partially fulfilled. %s\n",
> avail_pages < num_pages ? "Not enough memory." :
> "Balloon floor reached.");
>
> num_pages = avail_pages > floor ? (avail_pages - floor) : 0;
> - num_pages -= num_pages % PAGES_IN_2M;
> }
>
> while (!done) {
> @@ -1770,7 +1766,7 @@ static void balloon_up(union dm_msg_info *msg_info)
> num_ballooned = alloc_balloon_pages(&dm_device, num_pages,
> bl_resp, alloc_unit);
>
> - if (alloc_unit != 1 && num_ballooned == 0) {
> + if (alloc_unit != 1 && num_ballooned != num_pages) {
> alloc_unit = 1;
> continue;
> }

Thank you for addressing my comments on v1, this all looks good to me
now so:

Reviewed-by: Vitaly Kuznetsov <[email protected]>

--
Vitaly

2020-01-18 19:32:01

by Michael Kelley (LINUX)

[permalink] [raw]
Subject: RE: [PATCH V2] x86/Hyper-V: Balloon up according to request page number

From: Tianyu Lan <[email protected]> Sent: Thursday, January 16, 2020 6:16 AM
>
> Current code has assumption that balloon request memory size aligns
> with 2MB. But actually Hyper-V doesn't guarantee such alignment. When
> balloon driver receives non-aligned balloon request, it produces warning
> and balloon up more memory than requested in order to keep 2MB alignment.
> Remove the warning and balloon up memory according to actual requested
> memory size.
>
> Fixes: f6712238471a ("hv: hv_balloon: avoid memory leak on alloc_error of 2MB memory
> block")
> Cc: [email protected]
> Signed-off-by: Tianyu Lan <[email protected]>
> ---
> Change since v2:
> - Change logic of switching alloc_unit from 2MB to 4KB
> in the balloon_up() to avoid redundant iteration when
> handle non-aligned page request.
> - Remove 2MB alignment operation and comment in balloon_up()
> ---
> drivers/hv/hv_balloon.c | 12 ++++--------
> 1 file changed, 4 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/hv/hv_balloon.c b/drivers/hv/hv_balloon.c
> index 7f3e7ab22d5d..536807efbc35 100644
> --- a/drivers/hv/hv_balloon.c
> +++ b/drivers/hv/hv_balloon.c
> @@ -1684,7 +1684,7 @@ static unsigned int alloc_balloon_pages(struct
> hv_dynmem_device *dm,
> if (num_pages < alloc_unit)
> return 0;

The above test is no longer necessary. The num_pages < alloc_unit
case is handled implicitly by your new 'for' loop condition.

>
> - for (i = 0; (i * alloc_unit) < num_pages; i++) {
> + for (i = 0; i < num_pages / alloc_unit; i++) {
> if (bl_resp->hdr.size + sizeof(union dm_mem_page_range) >
> HV_HYP_PAGE_SIZE)
> return i * alloc_unit;
> @@ -1722,7 +1722,7 @@ static unsigned int alloc_balloon_pages(struct
> hv_dynmem_device *dm,
>
> }
>
> - return num_pages;
> + return i * alloc_unit;
> }
>
> static void balloon_up(union dm_msg_info *msg_info)
> @@ -1737,9 +1737,6 @@ static void balloon_up(union dm_msg_info *msg_info)
> long avail_pages;
> unsigned long floor;
>
> - /* The host balloons pages in 2M granularity. */
> - WARN_ON_ONCE(num_pages % PAGES_IN_2M != 0);
> -
> /*
> * We will attempt 2M allocations. However, if we fail to
> * allocate 2M chunks, we will go back to PAGE_SIZE allocations.
> @@ -1749,14 +1746,13 @@ static void balloon_up(union dm_msg_info *msg_info)
> avail_pages = si_mem_available();
> floor = compute_balloon_floor();
>
> - /* Refuse to balloon below the floor, keep the 2M granularity. */
> + /* Refuse to balloon below the floor. */
> if (avail_pages < num_pages || avail_pages - num_pages < floor) {
> pr_warn("Balloon request will be partially fulfilled. %s\n",
> avail_pages < num_pages ? "Not enough memory." :
> "Balloon floor reached.");
>
> num_pages = avail_pages > floor ? (avail_pages - floor) : 0;
> - num_pages -= num_pages % PAGES_IN_2M;
> }
>
> while (!done) {
> @@ -1770,7 +1766,7 @@ static void balloon_up(union dm_msg_info *msg_info)
> num_ballooned = alloc_balloon_pages(&dm_device, num_pages,
> bl_resp, alloc_unit);
>
> - if (alloc_unit != 1 && num_ballooned == 0) {
> + if (alloc_unit != 1 && num_ballooned != num_pages) {

Maybe I'm missing something, but I don't think Vitaly's optimization works. If
alloc_unit specifies 2 Mbytes, and num_pages specifies 3 Mbytes, then num_ballooned
will come back as 2 Mbytes, which is correct. But if we revert alloc_unit to 1 page and
"continue" in that case, we will lose the 2 Mbytes of memory (it's not freed), and the
next time through the loop will try to allocate only 1 Mbyte (because num_pages
will be decremented by num_ballooned). I think the original code does the right thing.

Michael

> alloc_unit = 1;
> continue;
> }
> --
> 2.14.5

2020-01-20 07:58:31

by Tianyu Lan

[permalink] [raw]
Subject: RE: [PATCH V2] x86/Hyper-V: Balloon up according to request page number

Hi Michael:
Thanks for your review.

> From: Michael Kelley <[email protected]>
> Sent: Sunday, January 19, 2020 3:29 AM
> To: [email protected]; KY Srinivasan <[email protected]>; Haiyang
> Zhang <[email protected]>; Stephen Hemminger
> <[email protected]>; [email protected]
> Cc: Tianyu Lan <[email protected]>; [email protected];
> [email protected]; vkuznets <[email protected]>;
> [email protected]
> Subject: RE: [PATCH V2] x86/Hyper-V: Balloon up according to request page
> number
>
> From: Tianyu Lan <[email protected]> Sent: Thursday, January 16,
> 2020 6:16 AM
> >
> > Current code has assumption that balloon request memory size aligns
> > with 2MB. But actually Hyper-V doesn't guarantee such alignment. When
> > balloon driver receives non-aligned balloon request, it produces
> > warning and balloon up more memory than requested in order to keep 2MB
> alignment.
> > Remove the warning and balloon up memory according to actual requested
> > memory size.
> >
> > Fixes: f6712238471a ("hv: hv_balloon: avoid memory leak on alloc_error
> > of 2MB memory
> > block")
> > Cc: [email protected]
> > Signed-off-by: Tianyu Lan <[email protected]>
> > ---
> > Change since v2:
> > - Change logic of switching alloc_unit from 2MB to 4KB
> > in the balloon_up() to avoid redundant iteration when
> > handle non-aligned page request.
> > - Remove 2MB alignment operation and comment in balloon_up()
> > ---
> > drivers/hv/hv_balloon.c | 12 ++++--------
> > 1 file changed, 4 insertions(+), 8 deletions(-)
> >
> > diff --git a/drivers/hv/hv_balloon.c b/drivers/hv/hv_balloon.c index
> > 7f3e7ab22d5d..536807efbc35 100644
> > --- a/drivers/hv/hv_balloon.c
> > +++ b/drivers/hv/hv_balloon.c
> > @@ -1684,7 +1684,7 @@ static unsigned int alloc_balloon_pages(struct
> > hv_dynmem_device *dm,
> > if (num_pages < alloc_unit)
> > return 0;
>
> The above test is no longer necessary. The num_pages < alloc_unit case is
> handled implicitly by your new 'for' loop condition.
>

Yes, will update in the next version.

> >
> > - for (i = 0; (i * alloc_unit) < num_pages; i++) {
> > + for (i = 0; i < num_pages / alloc_unit; i++) {
> > if (bl_resp->hdr.size + sizeof(union dm_mem_page_range) >
> > HV_HYP_PAGE_SIZE)
> > return i * alloc_unit;
> > @@ -1722,7 +1722,7 @@ static unsigned int alloc_balloon_pages(struct
> > hv_dynmem_device *dm,
> >
> > }
> >
> > - return num_pages;
> > + return i * alloc_unit;
> > }
> >
> > static void balloon_up(union dm_msg_info *msg_info) @@ -1737,9
> > +1737,6 @@ static void balloon_up(union dm_msg_info *msg_info)
> > long avail_pages;
> > unsigned long floor;
> >
> > - /* The host balloons pages in 2M granularity. */
> > - WARN_ON_ONCE(num_pages % PAGES_IN_2M != 0);
> > -
> > /*
> > * We will attempt 2M allocations. However, if we fail to
> > * allocate 2M chunks, we will go back to PAGE_SIZE allocations.
> > @@ -1749,14 +1746,13 @@ static void balloon_up(union dm_msg_info
> *msg_info)
> > avail_pages = si_mem_available();
> > floor = compute_balloon_floor();
> >
> > - /* Refuse to balloon below the floor, keep the 2M granularity. */
> > + /* Refuse to balloon below the floor. */
> > if (avail_pages < num_pages || avail_pages - num_pages < floor) {
> > pr_warn("Balloon request will be partially fulfilled. %s\n",
> > avail_pages < num_pages ? "Not enough memory." :
> > "Balloon floor reached.");
> >
> > num_pages = avail_pages > floor ? (avail_pages - floor) : 0;
> > - num_pages -= num_pages % PAGES_IN_2M;
> > }
> >
> > while (!done) {
> > @@ -1770,7 +1766,7 @@ static void balloon_up(union dm_msg_info
> *msg_info)
> > num_ballooned = alloc_balloon_pages(&dm_device,
> num_pages,
> > bl_resp, alloc_unit);
> >
> > - if (alloc_unit != 1 && num_ballooned == 0) {
> > + if (alloc_unit != 1 && num_ballooned != num_pages) {
>
> Maybe I'm missing something, but I don't think Vitaly's optimization works. If
> alloc_unit specifies 2 Mbytes, and num_pages specifies 3 Mbytes, then
> num_ballooned will come back as 2 Mbytes, which is correct. But if we revert
> alloc_unit to 1 page and "continue" in that case, we will lose the 2 Mbytes of
> memory (it's not freed), and the next time through the loop will try to allocate
> only 1 Mbyte (because num_pages will be decremented by num_ballooned). I
> think the original code does the right thing.
>
> Michael

Sorry. I should remove the "continue" here and then it will work. Will fix this in the
next version.

>
> > alloc_unit = 1;
> > continue;
> > }
> > --
> > 2.14.5

2020-01-20 10:22:54

by Vitaly Kuznetsov

[permalink] [raw]
Subject: RE: [PATCH V2] x86/Hyper-V: Balloon up according to request page number

Michael Kelley <[email protected]> writes:

> From: Tianyu Lan <[email protected]> Sent: Thursday, January 16, 2020 6:16 AM
>>
>> Current code has assumption that balloon request memory size aligns
>> with 2MB. But actually Hyper-V doesn't guarantee such alignment. When
>> balloon driver receives non-aligned balloon request, it produces warning
>> and balloon up more memory than requested in order to keep 2MB alignment.
>> Remove the warning and balloon up memory according to actual requested
>> memory size.
>>
>> Fixes: f6712238471a ("hv: hv_balloon: avoid memory leak on alloc_error of 2MB memory
>> block")
>> Cc: [email protected]
>> Signed-off-by: Tianyu Lan <[email protected]>
>> ---
>> Change since v2:
>> - Change logic of switching alloc_unit from 2MB to 4KB
>> in the balloon_up() to avoid redundant iteration when
>> handle non-aligned page request.
>> - Remove 2MB alignment operation and comment in balloon_up()
>> ---
>> drivers/hv/hv_balloon.c | 12 ++++--------
>> 1 file changed, 4 insertions(+), 8 deletions(-)
>>
>> diff --git a/drivers/hv/hv_balloon.c b/drivers/hv/hv_balloon.c
>> index 7f3e7ab22d5d..536807efbc35 100644
>> --- a/drivers/hv/hv_balloon.c
>> +++ b/drivers/hv/hv_balloon.c
>> @@ -1684,7 +1684,7 @@ static unsigned int alloc_balloon_pages(struct
>> hv_dynmem_device *dm,
>> if (num_pages < alloc_unit)
>> return 0;
>
> The above test is no longer necessary. The num_pages < alloc_unit
> case is handled implicitly by your new 'for' loop condition.
>
>>
>> - for (i = 0; (i * alloc_unit) < num_pages; i++) {
>> + for (i = 0; i < num_pages / alloc_unit; i++) {
>> if (bl_resp->hdr.size + sizeof(union dm_mem_page_range) >
>> HV_HYP_PAGE_SIZE)
>> return i * alloc_unit;
>> @@ -1722,7 +1722,7 @@ static unsigned int alloc_balloon_pages(struct
>> hv_dynmem_device *dm,
>>
>> }
>>
>> - return num_pages;
>> + return i * alloc_unit;
>> }
>>
>> static void balloon_up(union dm_msg_info *msg_info)
>> @@ -1737,9 +1737,6 @@ static void balloon_up(union dm_msg_info *msg_info)
>> long avail_pages;
>> unsigned long floor;
>>
>> - /* The host balloons pages in 2M granularity. */
>> - WARN_ON_ONCE(num_pages % PAGES_IN_2M != 0);
>> -
>> /*
>> * We will attempt 2M allocations. However, if we fail to
>> * allocate 2M chunks, we will go back to PAGE_SIZE allocations.
>> @@ -1749,14 +1746,13 @@ static void balloon_up(union dm_msg_info *msg_info)
>> avail_pages = si_mem_available();
>> floor = compute_balloon_floor();
>>
>> - /* Refuse to balloon below the floor, keep the 2M granularity. */
>> + /* Refuse to balloon below the floor. */
>> if (avail_pages < num_pages || avail_pages - num_pages < floor) {
>> pr_warn("Balloon request will be partially fulfilled. %s\n",
>> avail_pages < num_pages ? "Not enough memory." :
>> "Balloon floor reached.");
>>
>> num_pages = avail_pages > floor ? (avail_pages - floor) : 0;
>> - num_pages -= num_pages % PAGES_IN_2M;
>> }
>>
>> while (!done) {
>> @@ -1770,7 +1766,7 @@ static void balloon_up(union dm_msg_info *msg_info)
>> num_ballooned = alloc_balloon_pages(&dm_device, num_pages,
>> bl_resp, alloc_unit);
>>
>> - if (alloc_unit != 1 && num_ballooned == 0) {
>> + if (alloc_unit != 1 && num_ballooned != num_pages) {
>
> Maybe I'm missing something, but I don't think Vitaly's optimization works. If
> alloc_unit specifies 2 Mbytes, and num_pages specifies 3 Mbytes, then num_ballooned
> will come back as 2 Mbytes, which is correct. But if we revert alloc_unit to 1 page and
> "continue" in that case, we will lose the 2 Mbytes of memory (it's not freed), and the
> next time through the loop will try to allocate only 1 Mbyte (because num_pages
> will be decremented by num_ballooned). I think the original code does the right thing.
>

True, nice catch! What I meant to say is that we can avoid sending two
replies to the host and get away with just one. Unfortunately,
"num_ballooned != num_pages" modification is not sufficient as
alloc_balloon_pages() will overwrite bl_resp->range_array[] (as it
always starts from 0). I think we can still achieve the goal by
allocating bl_resp outside of the loop and filling it from
range_array[range_count] instead of range_array[i] but it seems to big
of a change for now particular gain. Let's just drop the idea.


> Michael
>
>> alloc_unit = 1;
>> continue;
>> }
>> --
>> 2.14.5
>

--
Vitaly