Skip to contentSkip to conversation

xen: patches for supporting sub-page and transitive grants

PATCH Superseded ANNIE LI 5 participants 16 messages Dec 6, 2011 – Dec 7, 2011 Xen hypervisor interface

v1 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

ANNIE LI

author
#0 lore raw

Hi

Following two patches introduce and implement interfaces for sub-page and transitive grants, and they are based on linux-next branch of linux/kernel/git/konrad/xen.git(3.2.0-rc2+). Sub-page and transitive grants are new types of grant table V2.

Descriptions for those patches:

Through sub-page grant table interfaces, a domain can grant another domain access to a range of bytes within a page, and Xen will then prevent the grantee domain accessing outside that range. For obvious reasons, it isn't possible to map these grant references, and domains are expected to use the grant copy hypercall instead.

Through transitive grant interfaces, a domain can create grant reference which redirects to another grant reference, so that any attempt to access the first grant reference will be redirected to the second one. This is used to implement receiver-side copy on inter-domain traffic: rather than copying the packet in dom0, dom0 creates a transitive grant referencing the original transmit buffer, and passes that to the receiving domain.

Diff:

 drivers/xen/grant-table.c |   74 
+++++++++++++++++++++++++++++++++++++++++++++
 include/xen/grant_table.h |   19 +++++++++++
 2 files changed, 93 insertions(+), 0 deletions(-)

Shortlog:

Annie Li (2):
xen/granttable: Support sub-page grants
xen/granttable: Support transitive grants

Thanks
Annie

#1 lore raw

PATCH 1/2 xen/granttable: Support sub-page grants

From: Annie Li <[email protected]>

-- They can't be used to map the page (so can only be used in a GNTTABOP_copy
hypercall).
-- It's possible to grant access with a finer granularity than whole pages. -- Xen guarantees that they can be revoked quickly (a normal map grant can
only be revoked with the cooperation of the domain which has been granted access).

---
 drivers/xen/grant-table.c |   41 +++++++++++++++++++++++++++++++++++++++++
 include/xen/grant_table.h |   13 +++++++++++++
 2 files changed, 54 insertions(+), 0 deletions(-)
drivers/xen/grant-table.c +41 -0
index bd325fd..7a0f4d1 100644
@@ -261,6 +261,47 @@ int gnttab_grant_foreign_access(domid_t domid, unsigned long frame,
 }
 EXPORT_SYMBOL_GPL(gnttab_grant_foreign_access);
 
+int gnttab_grant_foreign_access_subpage_v2(domid_t domid, unsigned long frame,
+					   int flags, unsigned page_off,
+					   unsigned length)
+{
+	int ref;
+
+	ref = get_free_entries(1);
+	if (unlikely(ref < 0))
+		return -ENOSPC;
+
+	gnttab_grant_foreign_access_ref_subpage_v2(ref, domid, frame, flags,
+						   page_off, length);
+
+	return ref;
+}
+EXPORT_SYMBOL_GPL(gnttab_grant_foreign_access_subpage_v2);
+
+void gnttab_grant_foreign_access_ref_subpage_v2(grant_ref_t ref, domid_t domid,
+						unsigned long frame, int flags,
+						unsigned page_off,
+						unsigned length)
+{
+	BUG_ON(flags & (GTF_accept_transfer | GTF_reading |
+			GTF_writing | GTF_transitive));
+	BUG_ON(grant_table_version == 1);
+	gnttab_shared.v2[ref].sub_page.frame = frame;
+	gnttab_shared.v2[ref].sub_page.page_off = page_off;
+	gnttab_shared.v2[ref].sub_page.length = length;
+	gnttab_shared.v2[ref].hdr.domid = domid;
+	wmb();
+	gnttab_shared.v2[ref].hdr.flags =
+				GTF_permit_access | GTF_sub_page | flags;
+}
+EXPORT_SYMBOL_GPL(gnttab_grant_foreign_access_ref_subpage_v2);
+
+bool gnttab_subpage_trans_grants_available(void)
+{
+	return grant_table_version == 2;
+}
+EXPORT_SYMBOL_GPL(gnttab_subpage_trans_grants_available);
+
 static int gnttab_query_foreign_access_v1(grant_ref_t ref)
 {
 	return gnttab_shared.v1[ref].flags & (GTF_reading|GTF_writing);
include/xen/grant_table.h +13 -0
index fea4954..7e43652 100644
@@ -62,6 +62,15 @@ int gnttab_resume(void);
 
 int gnttab_grant_foreign_access(domid_t domid, unsigned long frame,
 				int readonly);
+int gnttab_grant_foreign_access_subpage_v2(domid_t domid, unsigned long frame,
+					   int flags, unsigned page_off,
+					   unsigned length);
+
+/*
+ * Are sub-page or transitive grants available on this version of Xen?  Returns
+ * true if they are, and false if they're not.
+ */
+bool gnttab_subpage_trans_grants_available(void);
 
 /*
  * End access through the given grant reference, iff the grant entry is no
@@ -108,6 +117,10 @@ void gnttab_cancel_free_callback(struct gnttab_free_callback *callback);
 
 void gnttab_grant_foreign_access_ref(grant_ref_t ref, domid_t domid,
 				     unsigned long frame, int readonly);
+void gnttab_grant_foreign_access_ref_subpage_v2(grant_ref_t ref, domid_t domid,
+						unsigned long frame, int flags,
+						unsigned page_off,
+						unsigned length);
 
 void gnttab_grant_foreign_transfer_ref(grant_ref_t, domid_t domid,
 				       unsigned long pfn);
Signature · 1 line
1.7.6.4
#2 lore raw

Re: [PATCH 1/2] xen/granttable: Support sub-page grants

Quoted 9 lines · On Tue, 2011-12-06 at 10:56 +0000, [email protected] wrote:

From: Annie Li <[email protected]>

-- They can't be used to map the page (so can only be used in a GNTTABOP_copy
hypercall).
-- It's possible to grant access with a finer granularity than whole pages. -- Xen guarantees that they can be revoked quickly (a normal map grant can
only be revoked with the cooperation of the domain which has been granted access).

Almost all my comments will apply to the transitive grants patch too.

Quoted 11 lines
---
 drivers/xen/grant-table.c |   41 +++++++++++++++++++++++++++++++++++++++++
 include/xen/grant_table.h |   13 +++++++++++++
 2 files changed, 54 insertions(+), 0 deletions(-)
drivers/xen/grant-table.c +0 -0
index bd325fd..7a0f4d1 100644

drivers/xen/grant-table.c int gnttab_grant_foreign_access(domid_t domid, unsigned long frame, · line 266

Quoted 6 lines
drivers/xen/grant-table.c +3 -0
@@ -261,6 +261,47 @@ int gnttab_grant_foreign_access(domid_t domid, unsigned long frame,
 }
 EXPORT_SYMBOL_GPL(gnttab_grant_foreign_access);
 
+int gnttab_grant_foreign_access_subpage_v2(domid_t domid, unsigned long frame,
+					   int flags, unsigned page_off,
+					   unsigned length)

Please drop the v2 suffixes on the public functions.

Any reason not to route these via the ops table for consistency with all the other ops? Then your availability check becomes a test for NULL fn pointer rather than a specific version.

Quoted 12 lines
drivers/xen/grant-table.c +12 -0
+{
+	int ref;
+
+	ref = get_free_entries(1);
+	if (unlikely(ref < 0))
+		return -ENOSPC;
+
+	gnttab_grant_foreign_access_ref_subpage_v2(ref, domid, frame, flags,
+						   page_off, length);
+
+	return ref;
+}

drivers/xen/grant-table.c int gnttab_grant_foreign_access(domid_t domid, unsigned long frame, · line 288

Quoted 10 lines
drivers/xen/grant-table.c +10 -0
+EXPORT_SYMBOL_GPL(gnttab_grant_foreign_access_subpage_v2);
+
+void gnttab_grant_foreign_access_ref_subpage_v2(grant_ref_t ref, domid_t domid,
+						unsigned long frame, int flags,
+						unsigned page_off,
+						unsigned length)
+{
+	BUG_ON(flags & (GTF_accept_transfer | GTF_reading |
+			GTF_writing | GTF_transitive));
+	BUG_ON(grant_table_version == 1);

Returning -Esomething might be less drastic? ENOSYS perhaps?

Quoted 4 lines
drivers/xen/grant-table.c +4 -0
+	gnttab_shared.v2[ref].sub_page.frame = frame;
+	gnttab_shared.v2[ref].sub_page.page_off = page_off;
+	gnttab_shared.v2[ref].sub_page.length = length;
+	gnttab_shared.v2[ref].hdr.domid = domid;

drivers/xen/grant-table.c int gnttab_grant_foreign_access(domid_t domid, unsigned long frame, · line 302

Quoted 10 lines
drivers/xen/grant-table.c +10 -0
+	wmb();
+	gnttab_shared.v2[ref].hdr.flags =
+				GTF_permit_access | GTF_sub_page | flags;
+}
+EXPORT_SYMBOL_GPL(gnttab_grant_foreign_access_ref_subpage_v2);
+
+bool gnttab_subpage_trans_grants_available(void)
+{
+	return grant_table_version == 2;
+}

It's not clear this adds anything over and above letting the user query the grant table version. It's hard to tell since there are no users presented here though. Perhaps separate subpage and transitive functions would be cleaner?

Quoted 37 lines
drivers/xen/grant-table.c +2 -0
+EXPORT_SYMBOL_GPL(gnttab_subpage_trans_grants_available);
+
 static int gnttab_query_foreign_access_v1(grant_ref_t ref)
 {
 	return gnttab_shared.v1[ref].flags & (GTF_reading|GTF_writing);
include/xen/grant_table.h +9 -0
index fea4954..7e43652 100644
@@ -62,6 +62,15 @@ int gnttab_resume(void);
 
 int gnttab_grant_foreign_access(domid_t domid, unsigned long frame,
 				int readonly);
+int gnttab_grant_foreign_access_subpage_v2(domid_t domid, unsigned long frame,
+					   int flags, unsigned page_off,
+					   unsigned length);
 
+
+/*
+ * Are sub-page or transitive grants available on this version of Xen?  Returns
+ * true if they are, and false if they're not.
+ */
+bool gnttab_subpage_trans_grants_available(void);
 
 /*

* End access through the given grant reference, iff the grant entry is no

include/xen/grant_table.h +4 -0
@@ -108,6 +117,10 @@ void gnttab_cancel_free_callback(struct gnttab_free_callback *callback);
 
 void gnttab_grant_foreign_access_ref(grant_ref_t ref, domid_t domid,
 				     unsigned long frame, int readonly);
+void gnttab_grant_foreign_access_ref_subpage_v2(grant_ref_t ref, domid_t domid,
+						unsigned long frame, int flags,
+						unsigned page_off,
+						unsigned length);
 
 void gnttab_grant_foreign_transfer_ref(grant_ref_t, domid_t domid,
 				       unsigned long pfn);
#3 lore raw

Re: [PATCH 1/2] xen/granttable: Support sub-page grants

Quoted 15 lines · Ian Campbell
Quoted 15 lines
drivers/xen/grant-table.c +15 -0
+{
+	int ref;
+
+	ref = get_free_entries(1);
+	if (unlikely(ref < 0))
+		return -ENOSPC;
+
+	gnttab_grant_foreign_access_ref_subpage_v2(ref, domid, frame, flags,
+						   page_off, length);
+
+	return ref;
+}
+EXPORT_SYMBOL_GPL(gnttab_grant_foreign_access_subpage_v2);
+
+void gnttab_grant_foreign_access_ref_subpage_v2(grant_ref_t ref, domid_t domid,
Quoted 8 lines

drivers/xen/grant-table.c int gnttab_grant_foreign_access(domid_t domid, unsigned long frame, · line 288

Quoted 7 lines
drivers/xen/grant-table.c +7 -0
+						unsigned long frame, int flags,
+						unsigned page_off,
+						unsigned length)
+{
+	BUG_ON(flags & (GTF_accept_transfer | GTF_reading |
+			GTF_writing | GTF_transitive));
+	BUG_ON(grant_table_version == 1);

Returning -Esomething might be less drastic? ENOSYS perhaps?

Yeah, BUG_ON shouldn't be used for API misuse unless there's absolutely no other way to handle it.

Quoted 11 lines
Quoted 11 lines
drivers/xen/grant-table.c +10 -0
+	gnttab_shared.v2[ref].sub_page.frame = frame;
+	gnttab_shared.v2[ref].sub_page.page_off = page_off;
+	gnttab_shared.v2[ref].sub_page.length = length;
+	gnttab_shared.v2[ref].hdr.domid = domid;
+	wmb();
+	gnttab_shared.v2[ref].hdr.flags =
+				GTF_permit_access | GTF_sub_page | flags;
+}
+EXPORT_SYMBOL_GPL(gnttab_grant_foreign_access_ref_subpage_v2);
+
Quoted 8 lines

drivers/xen/grant-table.c int gnttab_grant_foreign_access(domid_t domid, unsigned long frame, · line 302

Quoted 4 lines
drivers/xen/grant-table.c +4 -0
+bool gnttab_subpage_trans_grants_available(void)
+{
+	return grant_table_version == 2;
+}

It's not clear this adds anything over and above letting the user query the grant table version. It's hard to tell since there are no users presented here though. Perhaps separate subpage and transitive functions would be cleaner?

Well, in general, specifically testing for features rather than interface versions is better.

J

author
#4 lore raw

Re: [PATCH 1/2] xen/granttable: Support sub-page grants

Quoted 5 lines
Quoted 3 lines

Returning -Esomething might be less drastic? ENOSYS perhaps?

Yeah, BUG_ON shouldn't be used for API misuse unless there's absolutely no other way to handle it.

OK, will change it, thanks.

Quoted 18 lines
Quoted 10 lines
Quoted 10 lines
drivers/xen/grant-table.c +10 -0
+	gnttab_shared.v2[ref].sub_page.frame = frame;
+	gnttab_shared.v2[ref].sub_page.page_off = page_off;
+	gnttab_shared.v2[ref].sub_page.length = length;
+	gnttab_shared.v2[ref].hdr.domid = domid;
+	wmb();
+	gnttab_shared.v2[ref].hdr.flags =
+				GTF_permit_access | GTF_sub_page | flags;
+}
+EXPORT_SYMBOL_GPL(gnttab_grant_foreign_access_ref_subpage_v2);
+
Quoted 8 lines

drivers/xen/grant-table.c int gnttab_grant_foreign_access(domid_t domid, unsigned long frame, · line 302

Quoted 4 lines
drivers/xen/grant-table.c +4 -0
+bool gnttab_subpage_trans_grants_available(void)
+{
+	return grant_table_version == 2;
+}

It's not clear this adds anything over and above letting the user query the grant table version. It's hard to tell since there are no users presented here though. Perhaps separate subpage and transitive functions would be cleaner?

Quoted 2 lines

Well, in general, specifically testing for features rather than interface versions is better.

Ok, will split it into 2 functions and check the function pointer instead of grant table version, thanks.

Thanks
Annie

author
#5 lore raw

Re: [PATCH 1/2] xen/granttable: Support sub-page grants

In reply to Ian Campbell

Thanks for your reviewing, Ian.

Quoted 6 lines

drivers/xen/grant-table.c int gnttab_grant_foreign_access(domid_t domid, unsigned long frame, · line 266

Quoted 5 lines
  EXPORT_SYMBOL_GPL(gnttab_grant_foreign_access);
drivers/xen/grant-table.c +3 -0
+int gnttab_grant_foreign_access_subpage_v2(domid_t domid, unsigned long frame,
+					   int flags, unsigned page_off,
+					   unsigned length)

Please drop the v2 suffixes on the public functions.

OK, the initial interface is without v2 suffixes. It was added in order to reminder user the interfaces are only available for grant table v2. But I am fine to remove it, and following ops fn pointers are better.

Quoted 3 lines

Any reason not to route these via the ops table for consistency with all the other ops? Then your availability check becomes a test for NULL fn pointer rather than a specific version.

Ok, it is good.
How about following implements?

gnttab_v1_ops = {
...
.access_subpage = NULL;
.access_ref_subpage = NULL;
.access_trans = NULL;
.access_ref_trans = NULL;
}

gnttab_v2_ops = {
...
.access_subpage = access_subpage_v2;
.access_ref_subpage = access_ref_subpage_v2;
.access_trans = access_trans_v2;
.access_ref_trans = access_ref_trans_v2;
}

gnttab_request_version()
{
.....
if(v2)

      gnttab_interface = &gnttab_v2_ops;
   else
      gnttab_interface = &gnttab_v1_ops;

.....
}

int gnttab_grant_foreign_access_subpage()

{
      if(gnttab_interface->access_subpage != NULL)
            return gnttab_interface->access_subpage;
      return Esomething;
}

Same operations for access_ref_subpage, access_trans and access_ref_trans.

bool gnttab_subpage_available()

{
     return (gnttab_interface->access_subpage != NULL);
}

bool gnttab_subpage_available()

{
     return (gnttab_interface->access_trans != NULL);
}

Thanks
Annie

#6 lore raw

RE: [PATCH 1/2] xen/granttable: Support sub-page grants

Quoted 39 lines

-----Original Message-----
From: ANNIE LI [mailto:[email protected]]
Sent: 07 December 2011 03:36
To: Ian Campbell
Cc: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; Paul Durrant
Subject: Re: [PATCH 1/2] xen/granttable: Support sub-page grants

Thanks for your reviewing, Ian.

Quoted 3 lines
Quoted 3 lines
  EXPORT_SYMBOL_GPL(gnttab_grant_foreign_access);

+int gnttab_grant_foreign_access_subpage_v2(domid_t domid,

unsigned long frame,

Quoted 3 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+					   int flags, unsigned page_off,
+					   unsigned length)

Please drop the v2 suffixes on the public functions.

OK, the initial interface is without v2 suffixes. It was added in order to reminder user the interfaces are only available for grant table v2.
But I am fine to remove it, and following ops fn pointers are better.

Quoted 1 line

Any reason not to route these via the ops table for consistency

with

Quoted 2 lines

all the other ops? Then your availability check becomes a test for NULL fn pointer rather than a specific version.

Ok, it is good.
How about following implements?

gnttab_v1_ops = {
...
.access_subpage = NULL;
.access_ref_subpage = NULL;
.access_trans = NULL;
.access_ref_trans = NULL;
}

Quoted 6 lines

gnttab_v2_ops = {
...
.access_subpage = access_subpage_v2; .access_ref_subpage = access_ref_subpage_v2; .access_trans = access_trans_v2; .access_ref_trans = access_ref_trans_v2; }

Do you need ops for the ref and non-ref functions? I would have thought you could just have the ref ones since the all the non-ref variants do is allocate and then call the ref variant.

Paul

Quoted 23 lines

gnttab_request_version()
{
.....

    if(v2)
       gnttab_interface = &gnttab_v2_ops;
    else
       gnttab_interface = &gnttab_v1_ops; .....
}

int gnttab_grant_foreign_access_subpage()

{
       if(gnttab_interface->access_subpage != NULL)
             return gnttab_interface->access_subpage;
       return Esomething;
}

Same operations for access_ref_subpage, access_trans and access_ref_trans.

bool gnttab_subpage_available()

{
      return (gnttab_interface->access_subpage != NULL); }
Quoted 6 lines

bool gnttab_subpage_available()

{
      return (gnttab_interface->access_trans != NULL); }

Thanks
Annie

���{.n�+�������+%�����ݶ��w��{.n�+����{��G�����{ay�ʇڙ�,j��f���h��������z_�(�階�ݢj"���m������G������?����&���~��iO��z��v�^�m���� ������?�I�

#7 lore raw

RE: [PATCH 1/2] xen/granttable: Support sub-page grants

Quoted 11 lines · Paul Durrant
Quoted 6 lines

gnttab_v2_ops = {
...
.access_subpage = access_subpage_v2; .access_ref_subpage = access_ref_subpage_v2; .access_trans = access_trans_v2; .access_ref_trans = access_ref_trans_v2; }

Do you need ops for the ref and non-ref functions? I would have thought you could just have the ref ones since the all the non-ref variants do is allocate and then call the ref variant.

Good point. It appears that all the existing ops are only present in the ref form.

Ian.

author
#8 lore raw

Re: [PATCH 1/2] xen/granttable: Support sub-page grants

Quoted 8 lines · Ian Campbell
Quoted 7 lines · Paul Durrant
Quoted 6 lines
gnttab_v2_ops = {
   ...

.access_subpage = access_subpage_v2; .access_ref_subpage = access_ref_subpage_v2; .access_trans = access_trans_v2; .access_ref_trans = access_ref_trans_v2; }

Quoted 5 lines
Quoted 3 lines

Do you need ops for the ref and non-ref functions? I would have thought you could just have the ref ones since the all the non-ref variants do is allocate and then call the ref variant.

Good point. It appears that all the existing ops are only present in the ref form.

So I will add two elements: access_ref_subpage and access_ref_trans into gnttab_v2_ops.

Thanks
Annie

Quoted 3 lines

Ian.

#9 lore raw

Re: [PATCH 1/2] xen/granttable: Support sub-page grants

Quoted 10 lines · ANNIE LI
Quoted 8 lines · Ian Campbell
Quoted 7 lines · Paul Durrant
Quoted 6 lines
gnttab_v2_ops = {
   ...

.access_subpage = access_subpage_v2; .access_ref_subpage = access_ref_subpage_v2; .access_trans = access_trans_v2; .access_ref_trans = access_ref_trans_v2; }

Quoted 7 lines
Quoted 5 lines
Quoted 3 lines

Do you need ops for the ref and non-ref functions? I would have thought you could just have the ref ones since the all the non-ref variants do is allocate and then call the ref variant.

Good point. It appears that all the existing ops are only present in the ref form.

So I will add two elements: access_ref_subpage and access_ref_trans into gnttab_v2_ops.

The existing convention seems to be for _ref to be a suffix, although it's only actually used on the end_*_ref ones.
+ int (*end_foreign_access_ref)(grant_ref_t, int);
+ unsigned long (*end_foreign_transfer_ref)(grant_ref_t);

Also it occurs to me that access_* sounds like something which uses a ref rather than something which sets one up. The existing hook to setup a normal grant is called "update_entry". Perhaps update_{subpage,transitive}_entry?

Ian.

author
#10 lore raw

Re: [PATCH 1/2] xen/granttable: Support sub-page grants

Quoted 10 lines

The existing convention seems to be for _ref to be a suffix, although it's only actually used on the end_*_ref ones.

(quoted diff) +2 -0
+       int (*end_foreign_access_ref)(grant_ref_t, int);
+       unsigned long (*end_foreign_transfer_ref)(grant_ref_t);

Also it occurs to me that access_* sounds like something which uses a ref rather than something which sets one up. The existing hook to setup a normal grant is called "update_entry". Perhaps update_{subpage,transitive}_entry?

Yes, you are right.
Just like the existing code:
gnttab_grant_foreign_access VS gnttab_grant_foreign_access_subpage
update_entry in gnttab_grant_foreign_access_ref VS update_{subpage,transitive}_entry in gnttab_grant_foreign_access_{subpage,trans}_ref

Thanks
Annie

#11 lore raw

Re: [Xen-devel] [PATCH 1/2] xen/granttable: Support sub-page grants

In reply to Paul Durrant
Quoted 8 lines
Quoted 8 lines

gnttab_v1_ops = {
...
.access_subpage = NULL;
.access_ref_subpage = NULL;
.access_trans = NULL;
.access_ref_trans = NULL;
}

You don't really need that. If you just leave it empty, they are all set to NULL.

#12 lore raw

Re: [PATCH 1/2] xen/granttable: Support sub-page grants

In reply to ANNIE LI
Quoted 7 lines · ANNIE LI

Thanks for your reviewing, Ian.

Quoted 6 lines

drivers/xen/grant-table.c int gnttab_grant_foreign_access(domid_t domid, unsigned long frame, · line 266

Quoted 5 lines
  EXPORT_SYMBOL_GPL(gnttab_grant_foreign_access);
drivers/xen/grant-table.c +3 -0
+int gnttab_grant_foreign_access_subpage_v2(domid_t domid, unsigned long frame,
+					   int flags, unsigned page_off,
+					   unsigned length)

Please drop the v2 suffixes on the public functions.

Quoted 8 lines

OK, the initial interface is without v2 suffixes. It was added in order to reminder user the interfaces are only available for grant table v2. But I am fine to remove it, and following ops fn pointers are better.

Quoted 3 lines

Any reason not to route these via the ops table for consistency with all the other ops? Then your availability check becomes a test for NULL fn pointer rather than a specific version.

Ok, it is good.
How about following implements?

Looks to be along the right lines. Thanks.

Quoted 7 lines

gnttab_v1_ops = {
...
.access_subpage = NULL;
.access_ref_subpage = NULL;
.access_trans = NULL;
.access_ref_trans = NULL;
}

I think you can omit these since NULL is the default but perhaps explicitly listing them is useful in a self documenting type way.

[...]

Quoted 9 lines

Same operations for access_ref_subpage, access_trans and access_ref_trans.

bool gnttab_subpage_available()

{
      return (gnttab_interface->access_subpage != NULL);
}

bool gnttab_subpage_available()

Typo: ..trans..

Quoted 3 lines
{
      return (gnttab_interface->access_trans != NULL);
}

Ian.

Quoted 3 lines

Thanks
Annie

author
#13 lore raw

Re: [PATCH 1/2] xen/granttable: Support sub-page grants

Quoted 25 lines · Ian Campbell
Quoted 7 lines · ANNIE LI

Thanks for your reviewing, Ian.

Quoted 6 lines

drivers/xen/grant-table.c int gnttab_grant_foreign_access(domid_t domid, unsigned long frame, · line 266

Quoted 5 lines
   EXPORT_SYMBOL_GPL(gnttab_grant_foreign_access);
drivers/xen/grant-table.c +3 -0
+int gnttab_grant_foreign_access_subpage_v2(domid_t domid, unsigned long frame,
+					   int flags, unsigned page_off,
+					   unsigned length)

Please drop the v2 suffixes on the public functions.

Quoted 8 lines

OK, the initial interface is without v2 suffixes. It was added in order to reminder user the interfaces are only available for grant table v2. But I am fine to remove it, and following ops fn pointers are better.

Quoted 3 lines

Any reason not to route these via the ops table for consistency with all the other ops? Then your availability check becomes a test for NULL fn pointer rather than a specific version.

Ok, it is good.
How about following implements?

Looks to be along the right lines. Thanks.

Quoted 7 lines
gnttab_v1_ops = {
   ...

.access_subpage = NULL;
.access_ref_subpage = NULL;
.access_trans = NULL;
.access_ref_trans = NULL;
}

Quoted 4 lines

I think you can omit these since NULL is the default but perhaps explicitly listing them is useful in a self documenting type way.

[...]

OK, I can delete those.

Quoted 9 lines
Quoted 8 lines

Same operations for access_ref_subpage, access_trans and access_ref_trans.

bool gnttab_subpage_available()

{
       return (gnttab_interface->access_subpage != NULL);
}

bool gnttab_subpage_available()

Typo: ..trans..

Thanks for pointing out this.

Thanks
Annie

Quoted 8 lines
Quoted 3 lines
{
       return (gnttab_interface->access_trans != NULL);
}

Ian.

Quoted 2 lines

Thanks
Annie

author
#14 lore raw

Re: [PATCH 1/2] xen/granttable: Support sub-page grants

In reply to [email protected]

drivers/xen/grant-table.c int gnttab_grant_foreign_access(domid_t domid, unsigned long frame, · line 287

Quoted 8 lines
drivers/xen/grant-table.c +7 -0
+void gnttab_grant_foreign_access_ref_subpage_v2(grant_ref_t ref, domid_t domid,
+						unsigned long frame, int flags,
+						unsigned page_off,
+						unsigned length)
+{
+	BUG_ON(flags&  (GTF_accept_transfer | GTF_reading |
+			GTF_writing | GTF_transitive));

This is slightly changed from the initial code, welcome your suggestions.

Initial condition is:
BUG_ON(flags & (GTF_accept_transfer | GTF_reading | GTF_writing | GTF_sub_page | GTF_permit_access));

GTF_sub_page | GTF_permit_access will be set later in this function, so it is necessary to verify this flag here. GTF_transitive and GTF_sub_page can not be enabled at same time, so GTF_transitive is checked here to avoid those two flags are both enabled.

Same code was changed in corresponding transitive function.

Thanks
Annie

Quoted 10 lines
drivers/xen/grant-table.c +9 -0
+	BUG_ON(grant_table_version == 1);
+	gnttab_shared.v2[ref].sub_page.frame = frame;
+	gnttab_shared.v2[ref].sub_page.page_off = page_off;
+	gnttab_shared.v2[ref].sub_page.length = length;
+	gnttab_shared.v2[ref].hdr.domid = domid;
+	wmb();
+	gnttab_shared.v2[ref].hdr.flags =
+				GTF_permit_access | GTF_sub_page | flags;
+}
#15 lore raw

PATCH 2/2 xen/granttable: Support transitive grants

In reply to ANNIE LI

From: Annie Li <[email protected]>

These allow a domain A which has been granted access on a page of domain B's memory to issue domain C with a copy-grant on the same page. This is useful e.g. for forwarding packets between domains.

---
 drivers/xen/grant-table.c |   33 +++++++++++++++++++++++++++++++++
 include/xen/grant_table.h |    6 ++++++
 2 files changed, 39 insertions(+), 0 deletions(-)
drivers/xen/grant-table.c +33 -0
index 7a0f4d1..d64b7c5 100644
@@ -296,6 +296,39 @@ void gnttab_grant_foreign_access_ref_subpage_v2(grant_ref_t ref, domid_t domid,
 }
 EXPORT_SYMBOL_GPL(gnttab_grant_foreign_access_ref_subpage_v2);
 
+int gnttab_grant_foreign_access_trans_v2(domid_t domid, int flags,
+					 domid_t trans_domid,
+					 grant_ref_t trans_gref)
+{
+	int ref;
+
+	ref = get_free_entries(1);
+	if (unlikely(ref < 0))
+		return -ENOSPC;
+
+	gnttab_grant_foreign_access_ref_trans_v2(ref, domid, flags,
+						 trans_domid, trans_gref);
+
+	return ref;
+}
+EXPORT_SYMBOL_GPL(gnttab_grant_foreign_access_trans_v2);
+
+void gnttab_grant_foreign_access_ref_trans_v2(grant_ref_t ref, domid_t domid,
+					      int flags, domid_t trans_domid,
+					      grant_ref_t trans_gref)
+{
+	BUG_ON(flags & (GTF_accept_transfer | GTF_reading |
+			GTF_writing | GTF_sub_page));
+	BUG_ON(grant_table_version == 1);
+	gnttab_shared.v2[ref].transitive.trans_domid = trans_domid;
+	gnttab_shared.v2[ref].transitive.gref = trans_gref;
+	gnttab_shared.v2[ref].hdr.domid = domid;
+	wmb();
+	gnttab_shared.v2[ref].hdr.flags =
+				GTF_permit_access | GTF_transitive | flags;
+}
+EXPORT_SYMBOL_GPL(gnttab_grant_foreign_access_ref_trans_v2);
+
 bool gnttab_subpage_trans_grants_available(void)
 {
 	return grant_table_version == 2;
include/xen/grant_table.h +6 -0
index 7e43652..35a8c73 100644
@@ -65,6 +65,9 @@ int gnttab_grant_foreign_access(domid_t domid, unsigned long frame,
 int gnttab_grant_foreign_access_subpage_v2(domid_t domid, unsigned long frame,
 					   int flags, unsigned page_off,
 					   unsigned length);
+int gnttab_grant_foreign_access_trans_v2(domid_t domid, int flags,
+					 domid_t trans_domid,
+					 grant_ref_t trans_gref);
 
 /*
  * Are sub-page or transitive grants available on this version of Xen?  Returns
@@ -121,6 +124,9 @@ void gnttab_grant_foreign_access_ref_subpage_v2(grant_ref_t ref, domid_t domid,
 						unsigned long frame, int flags,
 						unsigned page_off,
 						unsigned length);
+void gnttab_grant_foreign_access_ref_trans_v2(grant_ref_t ref, domid_t domid,
+					      int flags, domid_t trans_domid,
+					      grant_ref_t trans_gref);
 
 void gnttab_grant_foreign_transfer_ref(grant_ref_t, domid_t domid,
 				       unsigned long pfn);
Signature · 1 line
1.7.6.4