[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [RFC 1/3] xen/balloon: Allow allocating DMA buffers


  • To: Oleksandr Andrushchenko <andr2000@xxxxxxxxx>, "Oleksandr_Andrushchenko@xxxxxxxx" <Oleksandr_Andrushchenko@xxxxxxxx>, xen-devel@xxxxxxxxxxxxxxxxxxxx, linux-kernel@xxxxxxxxxxxxxxx, dri-devel@xxxxxxxxxxxxxxxxxxxxx, linux-media@xxxxxxxxxxxxxxx, jgross@xxxxxxxx, konrad.wilk@xxxxxxxxxx
  • From: Boris Ostrovsky <boris.ostrovsky@xxxxxxxxxx>
  • Date: Tue, 22 May 2018 15:09:17 -0400
  • Autocrypt: addr=boris.ostrovsky@xxxxxxxxxx; prefer-encrypt=mutual; keydata= xsFNBFH8CgsBEAC0KiOi9siOvlXatK2xX99e/J3OvApoYWjieVQ9232Eb7GzCWrItCzP8FUV PQg8rMsSd0OzIvvjbEAvaWLlbs8wa3MtVLysHY/DfqRK9Zvr/RgrsYC6ukOB7igy2PGqZd+M MDnSmVzik0sPvB6xPV7QyFsykEgpnHbvdZAUy/vyys8xgT0PVYR5hyvhyf6VIfGuvqIsvJw5 C8+P71CHI+U/IhsKrLrsiYHpAhQkw+Zvyeml6XSi5w4LXDbF+3oholKYCkPwxmGdK8MUIdkM d7iYdKqiP4W6FKQou/lC3jvOceGupEoDV9botSWEIIlKdtm6C4GfL45RD8V4B9iy24JHPlom woVWc0xBZboQguhauQqrBFooHO3roEeM1pxXjLUbDtH4t3SAI3gt4dpSyT3EvzhyNQVVIxj2 FXnIChrYxR6S0ijSqUKO0cAduenhBrpYbz9qFcB/GyxD+ZWY7OgQKHUZMWapx5bHGQ8bUZz2 SfjZwK+GETGhfkvNMf6zXbZkDq4kKB/ywaKvVPodS1Poa44+B9sxbUp1jMfFtlOJ3AYB0WDS Op3d7F2ry20CIf1Ifh0nIxkQPkTX7aX5rI92oZeu5u038dHUu/dO2EcuCjl1eDMGm5PLHDSP 0QUw5xzk1Y8MG1JQ56PtqReO33inBXG63yTIikJmUXFTw6lLJwARAQABzTNCb3JpcyBPc3Ry b3Zza3kgKFdvcmspIDxib3Jpcy5vc3Ryb3Zza3lAb3JhY2xlLmNvbT7CwXgEEwECACIFAlH8 CgsCGwMGCwkIBwMCBhUIAgkKCwQWAgMBAh4BAheAAAoJEIredpCGysGyasEP/j5xApopUf4g 9Fl3UxZuBx+oduuw3JHqgbGZ2siA3EA4bKwtKq8eT7ekpApn4c0HA8TWTDtgZtLSV5IdH+9z JimBDrhLkDI3Zsx2CafL4pMJvpUavhc5mEU8myp4dWCuIylHiWG65agvUeFZYK4P33fGqoaS VGx3tsQIAr7MsQxilMfRiTEoYH0WWthhE0YVQzV6kx4wj4yLGYPPBtFqnrapKKC8yFTpgjaK jImqWhU9CSUAXdNEs/oKVR1XlkDpMCFDl88vKAuJwugnixjbPFTVPyoC7+4Bm/FnL3iwlJVE qIGQRspt09r+datFzPqSbp5Fo/9m4JSvgtPp2X2+gIGgLPWp2ft1NXHHVWP19sPgEsEJXSr9 tskM8ScxEkqAUuDs6+x/ISX8wa5Pvmo65drN+JWA8EqKOHQG6LUsUdJolFM2i4Z0k40BnFU/ kjTARjrXW94LwokVy4x+ZYgImrnKWeKac6fMfMwH2aKpCQLlVxdO4qvJkv92SzZz4538az1T m+3ekJAimou89cXwXHCFb5WqJcyjDfdQF857vTn1z4qu7udYCuuV/4xDEhslUq1+GcNDjAhB nNYPzD+SvhWEsrjuXv+fDONdJtmLUpKs4Jtak3smGGhZsqpcNv8nQzUGDQZjuCSmDqW8vn2o hWwveNeRTkxh+2x1Qb3GT46uzsFNBFH8CgsBEADGC/yx5ctcLQlB9hbq7KNqCDyZNoYu1HAB Hal3MuxPfoGKObEktawQPQaSTB5vNlDxKihezLnlT/PKjcXC2R1OjSDinlu5XNGc6mnky03q yymUPyiMtWhBBftezTRxWRslPaFWlg/h/Y1iDuOcklhpr7K1h1jRPCrf1yIoxbIpDbffnuyz kuto4AahRvBU4Js4sU7f/btU+h+e0AcLVzIhTVPIz7PM+Gk2LNzZ3/on4dnEc/qd+ZZFlOQ4 KDN/hPqlwA/YJsKzAPX51L6Vv344pqTm6Z0f9M7YALB/11FO2nBB7zw7HAUYqJeHutCwxm7i BDNt0g9fhviNcJzagqJ1R7aPjtjBoYvKkbwNu5sWDpQ4idnsnck4YT6ctzN4I+6lfkU8zMzC gM2R4qqUXmxFIS4Bee+gnJi0Pc3KcBYBZsDK44FtM//5Cp9DrxRQOh19kNHBlxkmEb8kL/pw XIDcEq8MXzPBbxwHKJ3QRWRe5jPNpf8HCjnZz0XyJV0/4M1JvOua7IZftOttQ6KnM4m6WNIZ 2ydg7dBhDa6iv1oKdL7wdp/rCulVWn8R7+3cRK95SnWiJ0qKDlMbIN8oGMhHdin8cSRYdmHK kTnvSGJNlkis5a+048o0C6jI3LozQYD/W9wq7MvgChgVQw1iEOB4u/3FXDEGulRVko6xCBU4 SQARAQABwsFfBBgBAgAJBQJR/AoLAhsMAAoJEIredpCGysGyfvMQAIywR6jTqix6/fL0Ip8G jpt3uk//QNxGJE3ZkUNLX6N786vnEJvc1beCu6EwqD1ezG9fJKMl7F3SEgpYaiKEcHfoKGdh 30B3Hsq44vOoxR6zxw2B/giADjhmWTP5tWQ9548N4VhIZMYQMQCkdqaueSL+8asp8tBNP+TJ PAIIANYvJaD8xA7sYUXGTzOXDh2THWSvmEWWmzok8er/u6ZKdS1YmZkUy8cfzrll/9hiGCTj u3qcaOM6i/m4hqtvsI1cOORMVwjJF4+IkC5ZBoeRs/xW5zIBdSUoC8L+OCyj5JETWTt40+lu qoqAF/AEGsNZTrwHJYu9rbHH260C0KYCNqmxDdcROUqIzJdzDKOrDmebkEVnxVeLJBIhYZUd t3Iq9hdjpU50TA6sQ3mZxzBdfRgg+vaj2DsJqI5Xla9QGKD+xNT6v14cZuIMZzO7w0DoojM4 ByrabFsOQxGvE0w9Dch2BDSI2Xyk1zjPKxG1VNBQVx3flH37QDWpL2zlJikW29Ws86PHdthh Fm5PY8YtX576DchSP6qJC57/eAAe/9ztZdVAdesQwGb9hZHJc75B+VNm4xrh/PJO6c1THqdQ 19WVJ+7rDx3PhVncGlbAOiiiE3NOFPJ1OQYxPKtpBUukAlOTnkKE6QcA4zckFepUkfmBV1wM Jg6OxFYd01z+a+oL
  • Cc: daniel.vetter@xxxxxxxxx, matthew.d.roper@xxxxxxxxx, dongwon.kim@xxxxxxxxx
  • Delivery-date: Tue, 22 May 2018 19:06:30 +0000
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>
  • Openpgp: preference=signencrypt

On 05/22/2018 02:27 PM, Oleksandr Andrushchenko wrote:
> On 05/22/2018 09:02 PM, Boris Ostrovsky wrote:
>> On 05/22/2018 11:00 AM, Oleksandr Andrushchenko wrote:
>>> On 05/22/2018 05:33 PM, Boris Ostrovsky wrote:
>>>> On 05/22/2018 01:55 AM, Oleksandr Andrushchenko wrote:
>>>>> On 05/21/2018 11:36 PM, Boris Ostrovsky wrote:
>>>>>> On 05/21/2018 03:13 PM, Oleksandr Andrushchenko wrote:
>>>>>>> On 05/21/2018 09:53 PM, Boris Ostrovsky wrote:
>>>>>>>> On 05/21/2018 01:32 PM, Oleksandr Andrushchenko wrote:
>>>>>>>>> On 05/21/2018 07:35 PM, Boris Ostrovsky wrote:
>>>>>>>>>> On 05/21/2018 01:40 AM, Oleksandr Andrushchenko wrote:
>>>>>>>>>>> On 05/19/2018 01:04 AM, Boris Ostrovsky wrote:
>>>>>>>>>>>> On 05/17/2018 04:26 AM, Oleksandr Andrushchenko wrote:
>>>>>>>>>>>>> From: Oleksandr Andrushchenko
>>>>>>>>>>>>> <oleksandr_andrushchenko@xxxxxxxx>
>>>>>>>>>>>> A commit message would be useful.
>>>>>>>>>>> Sure, v1 will have it
>>>>>>>>>>>>> Signed-off-by: Oleksandr Andrushchenko
>>>>>>>>>>>>> <oleksandr_andrushchenko@xxxxxxxx>
>>>>>>>>>>>>>
>>>>>>>>>>>>>            for (i = 0; i < nr_pages; i++) {
>>>>>>>>>>>>> -        page = alloc_page(gfp);
>>>>>>>>>>>>> -        if (page == NULL) {
>>>>>>>>>>>>> -            nr_pages = i;
>>>>>>>>>>>>> -            state = BP_EAGAIN;
>>>>>>>>>>>>> -            break;
>>>>>>>>>>>>> +        if (ext_pages) {
>>>>>>>>>>>>> +            page = ext_pages[i];
>>>>>>>>>>>>> +        } else {
>>>>>>>>>>>>> +            page = alloc_page(gfp);
>>>>>>>>>>>>> +            if (page == NULL) {
>>>>>>>>>>>>> +                nr_pages = i;
>>>>>>>>>>>>> +                state = BP_EAGAIN;
>>>>>>>>>>>>> +                break;
>>>>>>>>>>>>> +            }
>>>>>>>>>>>>>                }
>>>>>>>>>>>>>                scrub_page(page);
>>>>>>>>>>>>>                list_add(&page->lru, &pages);
>>>>>>>>>>>>> @@ -529,7 +565,7 @@ static enum bp_state
>>>>>>>>>>>>> decrease_reservation(unsigned long nr_pages, gfp_t gfp)
>>>>>>>>>>>>>            i = 0;
>>>>>>>>>>>>>            list_for_each_entry_safe(page, tmp, &pages, lru) {
>>>>>>>>>>>>>                /* XENMEM_decrease_reservation requires a
>>>>>>>>>>>>> GFN */
>>>>>>>>>>>>> -        frame_list[i++] = xen_page_to_gfn(page);
>>>>>>>>>>>>> +        frames[i++] = xen_page_to_gfn(page);
>>>>>>>>>>>>>          #ifdef CONFIG_XEN_HAVE_PVMMU
>>>>>>>>>>>>>                /*
>>>>>>>>>>>>> @@ -552,18 +588,22 @@ static enum bp_state
>>>>>>>>>>>>> decrease_reservation(unsigned long nr_pages, gfp_t gfp)
>>>>>>>>>>>>>        #endif
>>>>>>>>>>>>>                list_del(&page->lru);
>>>>>>>>>>>>>        -        balloon_append(page);
>>>>>>>>>>>>> +        if (!ext_pages)
>>>>>>>>>>>>> +            balloon_append(page);
>>>>>>>>>>>> So what you are proposing is not really ballooning. You are
>>>>>>>>>>>> just
>>>>>>>>>>>> piggybacking on existing interfaces, aren't you?
>>>>>>>>>>> Sort of. Basically I need to
>>>>>>>>>>> {increase|decrease}_reservation, not
>>>>>>>>>>> actually
>>>>>>>>>>> allocating ballooned pages.
>>>>>>>>>>> Do you think I can simply EXPORT_SYMBOL for
>>>>>>>>>>> {increase|decrease}_reservation?
>>>>>>>>>>> Any other suggestion?
>>>>>>>>>> I am actually wondering how much of that code you end up
>>>>>>>>>> reusing.
>>>>>>>>>> You
>>>>>>>>>> pretty much create new code paths in both routines and common
>>>>>>>>>> code
>>>>>>>>>> ends
>>>>>>>>>> up being essentially the hypercall.
>>>>>>>>> Well, I hoped that it would be easier to maintain if I modify
>>>>>>>>> existing
>>>>>>>>> code
>>>>>>>>> to support both use-cases, but I am also ok to create new
>>>>>>>>> routines if
>>>>>>>>> this
>>>>>>>>> seems to be reasonable - please let me know
>>>>>>>>>>       So the question is --- would it make
>>>>>>>>>> sense to do all of this separately from the balloon driver?
>>>>>>>>> This can be done, but which driver will host this code then?
>>>>>>>>> If we
>>>>>>>>> move from
>>>>>>>>> the balloon driver, then this could go to either gntdev or
>>>>>>>>> grant-table.
>>>>>>>>> What's your preference?
>>>>>>>> A separate module?
>>>>>>>> Is there any use for this feature outside of your zero-copy DRM
>>>>>>>> driver?
>>>>>>> Intel's hyper dma-buf (Dongwon/Matt CC'ed), V4L/GPU at least.
>>>>>>>
>>>>>>> At the time I tried to upstream zcopy driver it was discussed and
>>>>>>> decided that
>>>>>>> it would be better if I remove all DRM specific code and move it to
>>>>>>> Xen drivers.
>>>>>>> Thus, this RFC.
>>>>>>>
>>>>>>> But it can also be implemented as a dedicated Xen dma-buf driver
>>>>>>> which
>>>>>>> will have all the
>>>>>>> code from this RFC + a bit more (char/misc device handling at
>>>>>>> least).
>>>>>>> This will also require a dedicated user-space library, just like
>>>>>>> libxengnttab.so
>>>>>>> for gntdev (now I have all new IOCTLs covered there).
>>>>>>>
>>>>>>> If the idea of a dedicated Xen dma-buf driver seems to be more
>>>>>>> attractive we
>>>>>>> can work toward this solution. BTW, I do support this idea, but
>>>>>>> was not
>>>>>>> sure if Xen community accepts yet another driver which duplicates
>>>>>>> quite some code
>>>>>>> of the existing gntdev/balloon/grant-table. And now after this
>>>>>>> RFC I
>>>>>>> hope that all cons
>>>>>>> and pros of both dedicated driver and gntdev/balloon/grant-table
>>>>>>> extension are
>>>>>>> clearly seen and we can make a decision.
>>>>>> IIRC the objection for a separate module was in the context of
>>>>>> gntdev
>>>>>> was discussion, because (among other things) people didn't want to
>>>>>> have
>>>>>> yet another file in /dev/xen/
>>>>>>
>>>>>> Here we are talking about (a new) balloon-like module which doesn't
>>>>>> create any new user-visible interfaces. And as for duplicating code
>>>>>> ---
>>>>>> as I said, I am not convinced there is much of duplication.
>>>>>>
>>>>>> I might even argue that we should add a new config option for this
>>>>>> module.
>>>>> I am not quite sure I am fully following you here: so, you suggest
>>>>> that we have balloon.c unchanged, but instead create a new
>>>>> module (namely a file under the same folder as balloon.c, e.g.
>>>>> dma-buf-reservation.c) and move those {increase|decrease}_reservation
>>>>> routines (specific to dma-buf) to that new file? And make it
>>>>> selectable
>>>>> via Kconfig? If so, then how about the changes to grant-table and
>>>>> gntdev?
>>>>> Those will look inconsistent then.
>>>> Inconsistent with what? The changes to grant code will also be
>>>> under the
>>>> new config option.
>>> Ah, ok.
>>>
>>> Option 1. We will have Kconfig option which will cover dma-buf
>>> changes in balloon,
>> I really don't think your changes to balloon driver belong there. The
>> have nothing to do with ballooning,
>>
>>> grant-table and gntdev. And for that we will
>>> create dedicated routines in balloon and grant-table (copy of
>>> the existing ones, but modified to fit dma-buf use-case) and
>>> those under something like "#if CONFIG_XEN_DMABUF"?
>>> This is relatively easy to do for balloon/grant-table, but not that
>>> easy for gntdev: there still seems to be lots of code which can be
>>> reused,
>>> so I'll have to put lots of "#if CONFIG_XEN_DMABUF" there. Even more,
>>> I change
>>> interfaces of the existing gntdev routines which won't look cute with
>>> #if's, IMO.
>>>
>>> Option 2. Try moving dma-buf related changes from balloon and
>>> grant-table to a new file. Then gntdev's Kconfig concerns from above
>>> will still
>>> be there, but balloon/grant-table functionality will be localized in a
>>> new module.
>> I don't see a problem with leaving your code (from patch 2) where it is
>> now, in grant table. It's a small change and it seems to me a single
>> #ifdef/#endif would cover it, even if you factor out common code there
>> as we've discussed. To my eye it logically belongs there. Just like your
>> gntdev changes belong to gntdev file. (Presumably, because I haven't
>> actually looked at them ;-))
>>
>> So my suggestion is
>> - separate module for your changes in balloon.c
> Ok, so, basically, the changes I need from the balloon driver is
> {increase|decrease}_reservation and DMAable memory allocations, so
> I'll move that into a separate file: what could be the name for such a
> file?


Naming would be your job ;-)


>
>> - keep grant-table changes, with config option
> Can we consider moving ex-balloon code into grant-table?

On the second thought ---  yes, if the code is compact enough, which I
think it is, you should be able to keep it there.


>
>> - keep gntdev changes, with config option.
> I'll try to see what happens to gntdev with Kconfig option wrt
> function prototype
> changes. I also have to check if UAPI of gntdev can also support
> CONFIG_XXX ifdefs
> w/o problems - do you by chance know if #if CONFIG_ is ok for UAPI files?


I would think that not but:

ostr@workbase> git grep "#ifdef CONFIG_" include/uapi/
include/uapi/asm-generic/mman-common.h:#ifdef
CONFIG_MMAP_ALLOW_UNINITIALIZED
include/uapi/linux/atmdev.h:#ifdef CONFIG_COMPAT
include/uapi/linux/elfcore.h:#ifdef CONFIG_BINFMT_ELF_FDPIC
include/uapi/linux/eventpoll.h:#ifdef CONFIG_PM_SLEEP
include/uapi/linux/fb.h:#ifdef CONFIG_FB_BACKLIGHT
include/uapi/linux/flat.h:#ifdef CONFIG_BINFMT_SHARED_FLAT
include/uapi/linux/hw_breakpoint.h:#ifdef CONFIG_HAVE_MIXED_BREAKPOINTS_REGS
ostr@workbase>


-boris


> Or I can leave UAPI as is and ifdef in .ioctl callback.
>>   (but when you get to post
>> actual patches I would appreciate if you could split this into a series
>> of logical changes and not post a one giant patch).
> Of course, as this is at RFC stage the idea was to roll out all the
> changes at once, so
> everyone has the full picture and don't need to collect changes from
> set of patches.
>>
>> -boris
>>
> Thank you,
> Oleksandr
>>> I am still missing your point here?
>>>
>>>>> If you suggest a new kernel driver module:
>>>>> IMO, there is nothing bad if we create a dedicated kernel module
>>>>> (driver) for Xen dma-buf handling selectable under Kconfig option.
>>>>> Yes, this will create a yet another device under /dev/xen,
>>>>> but most people will never see it if we set Kconfig to default to
>>>>> "n".
>>>>> And then we'll need user-space support for that, so Xen tools will
>>>>> be extended with libxendmabuf.so or so.
>>>>> This way all Xen dma-buf support can be localized at one place which
>>>>> might be easier to maintain. What is more it could be totally
>>>>> transparent
>>>>> to most of us as Kconfig option won't be set by default (both kernel
>>>>> and Xen).
>>>> The downside is that we will end up having another device for doing
>>>> things that are not that different from what we are already doing with
>>>> existing gnttab device. Or are they?
>>> Agree, but Kconfig option, IMO, won't make it look nice because
>>> of gntdev changes and code reuse.
>>>> -boris
>>> Thank you,
>>> Oleksandr
>>>
>>> _______________________________________________
>>> Xen-devel mailing list
>>> Xen-devel@xxxxxxxxxxxxxxxxxxxx
>>> https://lists.xenproject.org/mailman/listinfo/xen-devel
>


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.