[PATCH 1/2] ioctl_userfaultfd.2: add UFFD_FEATURE_EXACT_ADDRESS

[PATCH 1/2] ioctl_userfaultfd.2: add UFFD_FEATURE_EXACT_ADDRESS

[Nadav Amit]

From: Nadav Amit <namit@vmware.com>

Describe the new UFFD_FEATURE_EXACT_ADDRESS API feature.

Signed-off-by: Nadav Amit <namit@vmware.com>
---
 man2/ioctl_userfaultfd.2 | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/man2/ioctl_userfaultfd.2 b/man2/ioctl_userfaultfd.2
index 504f61d4b..94480158e 100644
--- a/man2/ioctl_userfaultfd.2
+++ b/man2/ioctl_userfaultfd.2
@@ -214,6 +214,14 @@ memory accesses to the regions registered with userfaultfd.
 If this feature bit is set,
 .I uffd_msg.pagefault.feat.ptid
 will be set to the faulted thread ID for each page-fault message.
+.TP
+.BR UFFD_FEATURE_EXACT_ADDRESS " (since Linux 5.18)"
+If this feature bit is set,
+.I uffd_msg.pagefault.address
+will be set to the exact page-fault address that was reported by the hardware,
+and will not mask the offset within the page.
+Note that old Linux versions might indicate the exact address as well,
+even though the feature bit is not set.
 .PP
 The returned
 .I ioctls
-- 
2.25.1

[PATCH 2/2] userfaultfd.2: fix userfaultfd_demo output

[Nadav Amit]

From: Nadav Amit <namit@vmware.com>

A bug in the kernel caused in recent version a different output (masked
offset). Update the output of the demo program accordingly.

Signed-off-by: Nadav Amit <namit@vmware.com>
---
 man2/userfaultfd.2 | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/man2/userfaultfd.2 b/man2/userfaultfd.2
index cee7c01d2..779ff8817 100644
--- a/man2/userfaultfd.2
+++ b/man2/userfaultfd.2
@@ -648,7 +648,7 @@ Address returned by mmap() = 0x7fd30106c000
 
 fault_handler_thread():
     poll() returns: nready = 1; POLLIN = 1; POLLERR = 0
-    UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fd30106c00f
+    UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fd30106c000
         (uffdio_copy.copy returned 4096)
 Read address 0x7fd30106c00f in main(): A
 Read address 0x7fd30106c40f in main(): A
@@ -657,7 +657,7 @@ Read address 0x7fd30106cc0f in main(): A
 
 fault_handler_thread():
     poll() returns: nready = 1; POLLIN = 1; POLLERR = 0
-    UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fd30106d00f
+    UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fd30106d000
         (uffdio_copy.copy returned 4096)
 Read address 0x7fd30106d00f in main(): B
 Read address 0x7fd30106d40f in main(): B
@@ -666,7 +666,7 @@ Read address 0x7fd30106dc0f in main(): B
 
 fault_handler_thread():
     poll() returns: nready = 1; POLLIN = 1; POLLERR = 0
-    UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fd30106e00f
+    UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fd30106e000
         (uffdio_copy.copy returned 4096)
 Read address 0x7fd30106e00f in main(): C
 Read address 0x7fd30106e40f in main(): C
-- 
2.25.1

Re: [PATCH 1/2] ioctl_userfaultfd.2: add UFFD_FEATURE_EXACT_ADDRESS

[Alejandro Colomar (man-pages)]

Hi Nadav,

On 3/7/22 19:48, Nadav Amit wrote:
> From: Nadav Amit <namit@vmware.com>
> 
> Describe the new UFFD_FEATURE_EXACT_ADDRESS API feature.
> 
> Signed-off-by: Nadav Amit <namit@vmware.com>
> ---

LGTM.  Please ping when it's added to Linus's tree,
and add a comment with the commit message that added it.

Thanks,

Alex

>  man2/ioctl_userfaultfd.2 | 8 ++++++++
>  1 file changed, 8 insertions(+)
> 
> diff --git a/man2/ioctl_userfaultfd.2 b/man2/ioctl_userfaultfd.2
> index 504f61d4b..94480158e 100644
> --- a/man2/ioctl_userfaultfd.2
> +++ b/man2/ioctl_userfaultfd.2
> @@ -214,6 +214,14 @@ memory accesses to the regions registered with userfaultfd.
>  If this feature bit is set,
>  .I uffd_msg.pagefault.feat.ptid
>  will be set to the faulted thread ID for each page-fault message.
> +.TP
> +.BR UFFD_FEATURE_EXACT_ADDRESS " (since Linux 5.18)"
> +If this feature bit is set,
> +.I uffd_msg.pagefault.address
> +will be set to the exact page-fault address that was reported by the hardware,
> +and will not mask the offset within the page.
> +Note that old Linux versions might indicate the exact address as well,
> +even though the feature bit is not set.
>  .PP
>  The returned
>  .I ioctls

-- 
Alejandro Colomar
Linux man-pages comaintainer; https://www.kernel.org/doc/man-pages/
http://www.alejandro-colomar.es/

Re: [PATCH 2/2] userfaultfd.2: fix userfaultfd_demo output

[Alejandro Colomar (man-pages)]

Hi Nadav,

On 3/7/22 19:48, Nadav Amit wrote:
> From: Nadav Amit <namit@vmware.com>
> 
> A bug in the kernel caused in recent version a different output (masked
> offset). Update the output of the demo program accordingly.

Please add to the commit message some info about the bug,
especially some links to LKML discussions and the patch set that fixes it.

Thanks,

Alex

> 
> Signed-off-by: Nadav Amit <namit@vmware.com>
> ---
>  man2/userfaultfd.2 | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/man2/userfaultfd.2 b/man2/userfaultfd.2
> index cee7c01d2..779ff8817 100644
> --- a/man2/userfaultfd.2
> +++ b/man2/userfaultfd.2
> @@ -648,7 +648,7 @@ Address returned by mmap() = 0x7fd30106c000
>  
>  fault_handler_thread():
>      poll() returns: nready = 1; POLLIN = 1; POLLERR = 0
> -    UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fd30106c00f
> +    UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fd30106c000
>          (uffdio_copy.copy returned 4096)
>  Read address 0x7fd30106c00f in main(): A
>  Read address 0x7fd30106c40f in main(): A
> @@ -657,7 +657,7 @@ Read address 0x7fd30106cc0f in main(): A
>  
>  fault_handler_thread():
>      poll() returns: nready = 1; POLLIN = 1; POLLERR = 0
> -    UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fd30106d00f
> +    UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fd30106d000
>          (uffdio_copy.copy returned 4096)
>  Read address 0x7fd30106d00f in main(): B
>  Read address 0x7fd30106d40f in main(): B
> @@ -666,7 +666,7 @@ Read address 0x7fd30106dc0f in main(): B
>  
>  fault_handler_thread():
>      poll() returns: nready = 1; POLLIN = 1; POLLERR = 0
> -    UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fd30106e00f
> +    UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fd30106e000
>          (uffdio_copy.copy returned 4096)
>  Read address 0x7fd30106e00f in main(): C
>  Read address 0x7fd30106e40f in main(): C

-- 
Alejandro Colomar
Linux man-pages comaintainer; https://www.kernel.org/doc/man-pages/
http://www.alejandro-colomar.es/

Re: [PATCH 1/2] ioctl_userfaultfd.2: add UFFD_FEATURE_EXACT_ADDRESS

[Alejandro Colomar]

[-- Attachment #1.1: Type: text/plain, Size: 1623 bytes --]

Hi Nadav,

On 3/14/22 13:23, Alejandro Colomar (man-pages) wrote:
> Hi Nadav,
> 
> On 3/7/22 19:48, Nadav Amit wrote:
>> From: Nadav Amit <namit@vmware.com>
>>
>> Describe the new UFFD_FEATURE_EXACT_ADDRESS API feature.
>>
>> Signed-off-by: Nadav Amit <namit@vmware.com>
>> ---
> 
> LGTM.  Please ping when it's added to Linus's tree,
> and add a comment with the commit message that added it.

I've checked that this was finally added to the kernel.  Should I apply
the patch as is, or is there anything that was changed?

Cheers,

Alex

> 
> Thanks,
> 
> Alex
> 
>>  man2/ioctl_userfaultfd.2 | 8 ++++++++
>>  1 file changed, 8 insertions(+)
>>
>> diff --git a/man2/ioctl_userfaultfd.2 b/man2/ioctl_userfaultfd.2
>> index 504f61d4b..94480158e 100644
>> --- a/man2/ioctl_userfaultfd.2
>> +++ b/man2/ioctl_userfaultfd.2
>> @@ -214,6 +214,14 @@ memory accesses to the regions registered with userfaultfd.
>>  If this feature bit is set,
>>  .I uffd_msg.pagefault.feat.ptid
>>  will be set to the faulted thread ID for each page-fault message.
>> +.TP
>> +.BR UFFD_FEATURE_EXACT_ADDRESS " (since Linux 5.18)"
>> +If this feature bit is set,
>> +.I uffd_msg.pagefault.address
>> +will be set to the exact page-fault address that was reported by the hardware,
>> +and will not mask the offset within the page.
>> +Note that old Linux versions might indicate the exact address as well,
>> +even though the feature bit is not set.
>>  .PP
>>  The returned
>>  .I ioctls
> 

-- 
<http://www.alejandro-colomar.es/>
GPG key fingerprint: A9348594CE31283A826FBDD8D57633D441E25BB5

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH 1/2] ioctl_userfaultfd.2: add UFFD_FEATURE_EXACT_ADDRESS

[Nadav Amit]

> On Mar 30, 2023, at 3:52 PM, Alejandro Colomar <alx.manpages@gmail.com> wrote:
> 
> Hi Nadav,
> 
> On 3/14/22 13:23, Alejandro Colomar (man-pages) wrote:
>> Hi Nadav,
>> 
>> On 3/7/22 19:48, Nadav Amit wrote:
>>> From: Nadav Amit <namit@vmware.com>
>>> 
>>> Describe the new UFFD_FEATURE_EXACT_ADDRESS API feature.
>>> 
>>> Signed-off-by: Nadav Amit <namit@vmware.com>
>>> ---
>> 
>> LGTM.  Please ping when it's added to Linus's tree,
>> and add a comment with the commit message that added it.
> 
> I've checked that this was finally added to the kernel.  Should I apply
> the patch as is, or is there anything that was changed?

Thanks for following up. Nothing to change, just apply it as is.

Regards,
Nadav

Re: [PATCH 1/2] ioctl_userfaultfd.2: add UFFD_FEATURE_EXACT_ADDRESS

[Alejandro Colomar]

[-- Attachment #1.1: Type: text/plain, Size: 978 bytes --]

Hi Nadav,

On 4/4/23 18:01, Nadav Amit wrote:
> 
> 
>> On Mar 30, 2023, at 3:52 PM, Alejandro Colomar <alx.manpages@gmail.com> wrote:
>>
>> Hi Nadav,
>>
>> On 3/14/22 13:23, Alejandro Colomar (man-pages) wrote:
>>> Hi Nadav,
>>>
>>> On 3/7/22 19:48, Nadav Amit wrote:
>>>> From: Nadav Amit <namit@vmware.com>
>>>>
>>>> Describe the new UFFD_FEATURE_EXACT_ADDRESS API feature.
>>>>
>>>> Signed-off-by: Nadav Amit <namit@vmware.com>
>>>> ---
>>>
>>> LGTM.  Please ping when it's added to Linus's tree,
>>> and add a comment with the commit message that added it.
>>
>> I've checked that this was finally added to the kernel.  Should I apply
>> the patch as is, or is there anything that was changed?
> 
> Thanks for following up. Nothing to change, just apply it as is.

Thanks!  I applied patch 1/2.

Cheers,
Alex

> 
> Regards,
> Nadav

-- 
<http://www.alejandro-colomar.es/>
GPG key fingerprint: A9348594CE31283A826FBDD8D57633D441E25BB5

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH 2/2] userfaultfd.2: fix userfaultfd_demo output

[Alejandro Colomar]

[-- Attachment #1.1: Type: text/plain, Size: 2167 bytes --]

Hi Nadav,

On 3/7/22 19:48, Nadav Amit wrote:
> From: Nadav Amit <namit@vmware.com>
> 
> A bug in the kernel caused in recent version a different output (masked
> offset). Update the output of the demo program accordingly.

I'd like to have some more information about this in the commit message.
I don't really understand it.  When was the bug introduced?  When was it
fixed?  Was the example in the page created in a buggy version?

Thanks,
Alex

> 
> Signed-off-by: Nadav Amit <namit@vmware.com>
> ---
>  man2/userfaultfd.2 | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/man2/userfaultfd.2 b/man2/userfaultfd.2
> index cee7c01d2..779ff8817 100644
> --- a/man2/userfaultfd.2
> +++ b/man2/userfaultfd.2
> @@ -648,7 +648,7 @@ Address returned by mmap() = 0x7fd30106c000
>  
>  fault_handler_thread():
>      poll() returns: nready = 1; POLLIN = 1; POLLERR = 0
> -    UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fd30106c00f
> +    UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fd30106c000
>          (uffdio_copy.copy returned 4096)
>  Read address 0x7fd30106c00f in main(): A
>  Read address 0x7fd30106c40f in main(): A
> @@ -657,7 +657,7 @@ Read address 0x7fd30106cc0f in main(): A
>  
>  fault_handler_thread():
>      poll() returns: nready = 1; POLLIN = 1; POLLERR = 0
> -    UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fd30106d00f
> +    UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fd30106d000
>          (uffdio_copy.copy returned 4096)
>  Read address 0x7fd30106d00f in main(): B
>  Read address 0x7fd30106d40f in main(): B
> @@ -666,7 +666,7 @@ Read address 0x7fd30106dc0f in main(): B
>  
>  fault_handler_thread():
>      poll() returns: nready = 1; POLLIN = 1; POLLERR = 0
> -    UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fd30106e00f
> +    UFFD_EVENT_PAGEFAULT event: flags = 0; address = 7fd30106e000
>          (uffdio_copy.copy returned 4096)
>  Read address 0x7fd30106e00f in main(): C
>  Read address 0x7fd30106e40f in main(): C

-- 
<http://www.alejandro-colomar.es/>
GPG key fingerprint: A9348594CE31283A826FBDD8D57633D441E25BB5

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH 2/2] userfaultfd.2: fix userfaultfd_demo output

[Nadav Amit]

> On Apr 8, 2023, at 6:34 AM, Alejandro Colomar <alx.manpages@gmail.com> wrote:
> 
> Hi Nadav,
> 
> On 3/7/22 19:48, Nadav Amit wrote:
>> From: Nadav Amit <namit@vmware.com>
>> 
>> A bug in the kernel caused in recent version a different output (masked
>> offset). Update the output of the demo program accordingly.
> 
> I'd like to have some more information about this in the commit message.
> I don't really understand it.  When was the bug introduced?  When was it
> fixed?  Was the example in the page created in a buggy version?

Hi Alex,

Sorry for the late response.

The bug was introduced in 1a29d85eb0f1 ("mm: use vmf->address instead of of
vmf->virtual_address”) which is circa 2016. It was introduced shortly after
userfaultfd, so it was determined that no users are likely to rely on the
intended correct behavior that was documented in the man page, but there
might be users who relied on the wrong behavior, which was in the code
for several years.

More information is available in the patch of the fix:

https://lore.kernel.org/all/20220226022655.350562-1-namit@vmware.com/

Let me know if it answers you questions, and whether you want me to send
an updated patch for the man page.

Re: [PATCH 2/2] userfaultfd.2: fix userfaultfd_demo output

[Alejandro Colomar]

[-- Attachment #1.1: Type: text/plain, Size: 1541 bytes --]

Hi!

On 4/10/23 21:57, Nadav Amit wrote:
> 
>> On Apr 8, 2023, at 6:34 AM, Alejandro Colomar <alx.manpages@gmail.com> wrote:
>>
>> Hi Nadav,
>>
>> On 3/7/22 19:48, Nadav Amit wrote:
>>> From: Nadav Amit <namit@vmware.com>
>>>
>>> A bug in the kernel caused in recent version a different output (masked
>>> offset). Update the output of the demo program accordingly.
>>
>> I'd like to have some more information about this in the commit message.
>> I don't really understand it.  When was the bug introduced?  When was it
>> fixed?  Was the example in the page created in a buggy version?
> 
> Hi Alex,
> 
> Sorry for the late response.
> 
> The bug was introduced in 1a29d85eb0f1 ("mm: use vmf->address instead of of
> vmf->virtual_address”) which is circa 2016. It was introduced shortly after
> userfaultfd, so it was determined that no users are likely to rely on the
> intended correct behavior that was documented in the man page, but there
> might be users who relied on the wrong behavior, which was in the code
> for several years.
> 
> More information is available in the patch of the fix:
> 
> https://lore.kernel.org/all/20220226022655.350562-1-namit@vmware.com/
> 
> Let me know if it answers you questions, and whether you want me to send
> an updated patch for the man page.

Yes, please, a patch including all this response in the message would be
useful.

Thanks!
Alex

> 

-- 
<http://www.alejandro-colomar.es/>
GPG key fingerprint: A9348594CE31283A826FBDD8D57633D441E25BB5

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]