Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
From:   William Kucharski <kucharsk@gmail.com>
Mime-Version: 1.0 (1.0)
Subject: Re: [RFC PATCH 0/6] Add support for shared PTEs across processes
Date:   Tue, 25 Jan 2022 05:09:03 -0700
Message-Id: <E44A9AB1-DBF0-4B8E-B049-293DD4DE6093@gmail.com>
References: <20220125114212.ks2qtncaahi6foan@box.shutemov.name>
Cc:     Khalid Aziz <khalid.aziz@oracle.com>, akpm@linux-foundation.org,
        willy@infradead.org, longpeng2@huawei.com, arnd@arndb.de,
        dave.hansen@linux.intel.com, david@redhat.com, rppt@kernel.org,
        surenb@google.com, linux-kernel@vger.kernel.org, linux-mm@kvack.org
In-Reply-To: <20220125114212.ks2qtncaahi6foan@box.shutemov.name>
To:     "Kirill A. Shutemov" <kirill@shutemov.name>
Precedence: bulk

I would think this should be the case; certainly it seems to be a more effec=
tive approach than having to manually enable sharing via the API in every ca=
se or via changes to ld.so.

If anything it might be useful to have an API for shutting it off, though th=
ere are already multiple areas where the system shares resources in ways tha=
t cannot be shut off by user action.

> On Jan 25, 2022, at 04:41, Kirill A. Shutemov <kirill@shutemov.name> wrote=
:
>=20
> =EF=BB=BFOn Tue, Jan 18, 2022 at 02:19:12PM -0700, Khalid Aziz wrote:
>> Example Code
>> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
>>=20
>> Snippet of the code that a donor process would run looks like below:
>>=20
>> -----------------
>>        addr =3D mmap((void *)TB(2), GB(512), PROT_READ | PROT_WRITE,
>>                        MAP_SHARED | MAP_ANONYMOUS, 0, 0);
>>        if (addr =3D=3D MAP_FAILED)
>>                perror("ERROR: mmap failed");
>>=20
>>        err =3D syscall(MSHARE_SYSCALL, "testregion", (void *)TB(2),
>>            GB(512), O_CREAT|O_RDWR|O_EXCL, 600);
>>        if (err < 0) {
>>                perror("mshare() syscall failed");
>>                exit(1);
>>        }
>>=20
>>        strncpy(addr, "Some random shared text",
>>            sizeof("Some random shared text"));
>> -----------------
>>=20
>> Snippet of code that a consumer process would execute looks like:
>>=20
>> -----------------
>>        fd =3D open("testregion", O_RDONLY);
>>        if (fd < 0) {
>>                perror("open failed");
>>                exit(1);
>>        }
>>=20
>>        if ((count =3D read(fd, &mshare_info, sizeof(mshare_info)) > 0))
>>                printf("INFO: %ld bytes shared at addr %lx \n",
>>                mshare_info[1], mshare_info[0]);
>>        else
>>                perror("read failed");
>>=20
>>        close(fd);
>>=20
>>        addr =3D (char *)mshare_info[0];
>>        err =3D syscall(MSHARE_SYSCALL, "testregion", (void *)mshare_info[=
0],
>>            mshare_info[1], O_RDWR, 600);
>>        if (err < 0) {
>>                perror("mshare() syscall failed");
>>                exit(1);
>>        }
>>=20
>>        printf("Guest mmap at %px:\n", addr);
>>        printf("%s\n", addr);
>>    printf("\nDone\n");
>>=20
>>        err =3D syscall(MSHARE_UNLINK_SYSCALL, "testregion");
>>        if (err < 0) {
>>                perror("mshare_unlink() failed");
>>                exit(1);
>>        }
>> -----------------
>=20
> I wounder if we can get away with zero-API here: we can transparently
> create/use shared page tables for any inode on mmap(MAP_SHARED) as long as=

> size and alignment is sutiable. Page tables will be linked to the inode
> and will be freed when the last of such mapping will go away. I don't see
> a need in new syscalls of flags to existing one.
>=20
> --=20
> Kirill A. Shutemov
>=20