Received-SPF: pass (google.com: domain of linux-kernel+bounces-173386-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Date: Wed, 8 May 2024 10:11:28 -0400
From: Peter Xu <peterx@redhat.com>
To: Audra Mitchell <audra@redhat.com>
Cc: David Hildenbrand <david@redhat.com>, viro@zeniv.linux.org.uk,
	brauner@kernel.org, jack@suse.cz, linux-fsdevel@vger.kernel.org,
	linux-kernel@vger.kernel.org, akpm@linux-foundation.org,
	shuah@kernel.org, linux-kselftest@vger.kernel.org,
	linux-mm@kvack.org, raquini@redhat.com
Subject: Re: [PATCH 1/2] Fix userfaultfd_api to return EINVAL as expected
Message-ID: <ZjuIEH8TW2tWcqXQ@x1n>
References: <20240507195510.283744-1-audra@redhat.com>
 <939a16f2-7b66-45a6-a043-4821bd3c71dc@redhat.com>
 <Zjt3Apr8ILFA4oK_@fedora>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <Zjt3Apr8ILFA4oK_@fedora>

On Wed, May 08, 2024 at 08:58:42AM -0400, Audra Mitchell wrote:
> On Wed, May 08, 2024 at 09:39:10AM +0200, David Hildenbrand wrote:
> > On 07.05.24 21:55, Audra Mitchell wrote:
> > > Currently if we request a feature that is not set in the Kernel
> > > config we fail silently and return the available features. However, the
> > > documentation indicates we should return an EINVAL.
> > 
> > I assume you are referencing
> > 
> > "EINVAL The API version requested in the api field is not supported by this
> > kernel, or  the  features  field passed to the kernel includes feature bits
> > that are not supported by the current kernel version."
> > 
> > and
> > 
> > "To  enable  userfaultfd features the application should set a bit
> > corresponding to each feature it wants to enable in the features field. If
> > the kernel supports all the requested features it will enable them.
> > Otherwise it will zero out the returned uffdio_api structure and return
> > EINVAL.
> > "
> > 
> > in which case I agree.
> 
> Yep! I'm referencing the man page.
> 
> > 
> > > 
> > > We need to fix this issue since we can end up with a Kernel warning
> > > should a program request the feature UFFD_FEATURE_WP_UNPOPULATED on
> > > a kernel with the config not set with this feature.
> > 
> > Can you mention which exact one? Is it a WARN* or a pr_warn() ?
> 
> Here is the kernel warning I get:
> 
> [  200.803094] unrecognized swap entry 0x7c00000000000001
> [  200.808270] ------------[ cut here ]------------
> [  200.812896] WARNING: CPU: 91 PID: 13634 at mm/memory.c:1660 zap_pte_range+0x43d/0x660
> [  200.820738] Modules linked in: qrtr bridge stp llc rfkill sunrpc amd_atl intel_rapl_msr intel_rapl_common amd64_edac edac_mce_amd kvm_amd kvm ipmi_ssif acpi_ipmi i2c_piix4 ipmi_si wmi_bmof dcdbas dell_smbios dell_wmi_descriptor ptdma ipmi_devintf rapl ipmi_msghandler acpi_power_meter pcspkr k10temp xfs libcrc32c sd_mod t10_pi mgag200 sg drm_kms_helper crct10dif_pclmul i2c_algo_bit ahci crc32_pclmul drm_shmem_helper libahci crc32c_intel drm i40e libata ghash_clmulni_intel tg3 ccp megaraid_sas sp5100_tco wmi dm_mirror dm_region_hash dm_log dm_mod fuse
> [  200.869387] CPU: 91 PID: 13634 Comm: userfaultfd Kdump: loaded Not tainted 6.9.0-rc5+ #8
> [  200.877477] Hardware name: Dell Inc. PowerEdge R6525/0N7YGH, BIOS 2.7.3 03/30/2022
> [  200.885052] RIP: 0010:zap_pte_range+0x43d/0x660
> [  200.889595] Code: 83 fa 02 0f 86 44 01 00 00 83 f9 17 0f 84 e1 00 00 00 83 f9 1f 0f 84 d0 00 00 00 48 89 c6 48 c7 c7 00 e4 dd bb e8 73 a2 de ff <0f> 0b e9 44 fd ff ff 45 0f b6 44 24 20 41 f6 c0 f4 75 27 4c 89 ee
> [  200.908348] RSP: 0018:ffffa18d2e6c37c8 EFLAGS: 00010246
> [  200.913584] RAX: 000000000000002a RBX: 00007f26d3600000 RCX: 0000000000000000
> [  200.920730] RDX: 0000000000000000 RSI: ffff93503f9a0bc0 RDI: ffff93503f9a0bc0
> [  200.927867] RBP: 00007f26d35cc000 R08: 0000000000000000 R09: ffffa18d2e6c3688
> [  200.935009] R10: ffffa18d2e6c3680 R11: ffffffffbc9de448 R12: ffffa18d2e6c39e8
> [  200.942149] R13: ffff92d1ebc15b50 R14: ffff93114e0cde60 R15: ffffa18d2e6c3928
> [  200.949291] FS:  0000000000000000(0000) GS:ffff93503f980000(0000) knlGS:0000000000000000
> [  200.957384] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [  200.963140] CR2: 00007f26b1600658 CR3: 00000040905ba000 CR4: 0000000000350ef0
> [  200.970283] Call Trace:
> [  200.972745]  <TASK>
> [  200.974862]  ? __warn+0x7f/0x130
> [  200.978108]  ? zap_pte_range+0x43d/0x660
> [  200.982044]  ? report_bug+0x18a/0x1a0
> [  200.985720]  ? handle_bug+0x3c/0x70
> [  200.989219]  ? exc_invalid_op+0x14/0x70
> [  200.993068]  ? asm_exc_invalid_op+0x16/0x20
> [  200.997265]  ? zap_pte_range+0x43d/0x660
> [  201.001199]  ? zap_pte_range+0x43d/0x660
> [  201.005134]  zap_pmd_range.isra.0+0xf9/0x230
> [  201.009416]  unmap_page_range+0x2d4/0x4a0
> [  201.013436]  unmap_vmas+0xa8/0x180
> [  201.016854]  exit_mmap+0xea/0x3b0
> [  201.020191]  __mmput+0x43/0x120
> [  201.023342]  exit_mm+0xb1/0x110
> [  201.026496]  do_exit+0x270/0x4f0
> [  201.029739]  do_group_exit+0x2c/0x80
> [  201.033326]  get_signal+0x886/0x8b0
> [  201.036828]  ? srso_return_thunk+0x5/0x5f
> [  201.040848]  arch_do_signal_or_restart+0x25/0x100
> [  201.045563]  ? srso_return_thunk+0x5/0x5f
> [  201.049583]  ? vma_set_page_prot+0x5e/0xc0
> [  201.053692]  ? srso_return_thunk+0x5/0x5f
> [  201.057713]  ? syscall_exit_work+0xff/0x130
> [  201.061908]  syscall_exit_to_user_mode+0x1b3/0x200
> [  201.066712]  do_syscall_64+0x87/0x160
> [  201.070387]  ? srso_return_thunk+0x5/0x5f
> [  201.074405]  ? do_mmap+0x416/0x5f0
> [  201.077821]  ? srso_return_thunk+0x5/0x5f
> [  201.081840]  ? rseq_get_rseq_cs+0x1d/0x240
> [  201.085950]  ? srso_return_thunk+0x5/0x5f
> [  201.089970]  ? rseq_ip_fixup+0x6d/0x1d0
> [  201.093823]  ? vm_mmap_pgoff+0x117/0x1a0
> [  201.097755]  ? srso_return_thunk+0x5/0x5f
> [  201.101776]  ? srso_return_thunk+0x5/0x5f
> [  201.105795]  ? syscall_exit_to_user_mode+0x78/0x200
> [  201.110685]  ? srso_return_thunk+0x5/0x5f
> [  201.114706]  ? do_syscall_64+0x87/0x160
> [  201.118557]  ? srso_return_thunk+0x5/0x5f
> [  201.122575]  ? __count_memcg_events+0x49/0xb0
> [  201.126944]  ? srso_return_thunk+0x5/0x5f
> [  201.130967]  ? srso_return_thunk+0x5/0x5f
> [  201.134986]  ? syscall_exit_work+0xff/0x130
> [  201.139184]  ? srso_return_thunk+0x5/0x5f
> [  201.143205]  ? syscall_exit_to_user_mode+0x78/0x200
> [  201.148093]  ? srso_return_thunk+0x5/0x5f
> [  201.152114]  ? do_syscall_64+0x87/0x160
> [  201.155960]  ? srso_return_thunk+0x5/0x5f
> [  201.159984]  ? sched_clock_cpu+0xb/0x190
> [  201.163916]  ? srso_return_thunk+0x5/0x5f
> [  201.167939]  ? irqtime_account_irq+0x40/0xc0
> [  201.172220]  ? srso_return_thunk+0x5/0x5f
> [  201.176243]  ? srso_return_thunk+0x5/0x5f
> [  201.180263]  entry_SYSCALL_64_after_hwframe+0x76/0x7e
> [  201.185326] RIP: 0033:0x7f26dfd0735b
> [  201.188939] Code: Unable to access opcode bytes at 0x7f26dfd07331.
> [  201.195128] RSP: 002b:00007fffce176868 EFLAGS: 00000206 ORIG_RAX: 000000000000000a
> [  201.202700] RAX: fffffffffffffffc RBX: 00007f26dfe60000 RCX: 00007f26dfd0735b
> [  201.209841] RDX: 0000000000000003 RSI: 0000000001000000 RDI: 00007f26af401000
> [  201.216983] RBP: 00007f26b0400640 R08: 00000000ffffffff R09: 0000000000000000
> [  201.224127] R10: ffffffffffffffc0 R11: 0000000000000206 R12: 0000000000000000
> [  201.231267] R13: 000000000040d320 R14: 0000000000000000 R15: 0000000000000000
> [  201.238413]  </TASK>
> [  201.240610] ---[ end trace 0000000000000000 ]---
> [  201.245250] unrecognized swap entry 0x7c00000000000001
> 
> 
> 
> > 
> > Likely we want "Fixes:" here.
> 
> This could be seen as a continuation of the problem 
> 2ff559f31a5d Revert "userfaultfd: don't fail on unrecognized features" 
> was trying to solve. However, this patch only checks to make sure we didnt 
> ask for a feature outside the possible range of features. We are still missing
> a check to confirm the requested features are also configured on. So I guess 
> the "Fixes" tag would be for this patch?
> 914eedcb9ba0 userfaultfd: don't fail on unrecognized features

Even though 914eedcb9ba0 was problematic but it might be innocent in this
regard, as we miss such check even before that commit.

Perhaps the right Fixes should be when there will be flags that will
only be supported conditionally, which should be the wp support.  In that
case, it could be:

Fixes: e06f1e1dd499 ("userfaultfd: wp: enabled write protection in userfaultfd API")

But it's still complicated though even if we want to backport, as this
specific check for userfault was added later... in:

00b151f21f39 ("mm/userfaultfd: fail uffd-wp registration if not supported")

In summary: I think we can stick with Fixes on e06f1e1dd499, but we don't
copy stable.  The major reason we don't copy stable here is not only about
complexity of such backport, but also that there can be apps trying to pass
in unsupported bits (even if the kernel didn't support it) but keep using
MISSING mode only, then we shouldn't fail them easily after a stable
upgrade.  Just smells dangerous to backport.

When at it and repost, would you mind consolidate the two feature list
check together?  We had the previous check here:

	if (uffdio_api.api != UFFD_API || (features & ~UFFD_API_FEATURES))
		goto err_out;

If with this patch IIUC this previous check can be removed (still need to
check UFFD_API).

Thanks,

> 
> Happy to get your input here!
> 
> Thanks in advance!
> 
> 
> > 
> > > 
> > > Signed-off-by: Audra Mitchell <audra@redhat.com>
> > > ---
> > >   fs/userfaultfd.c | 5 +++++
> > >   1 file changed, 5 insertions(+)
> > > 
> > > diff --git a/fs/userfaultfd.c b/fs/userfaultfd.c
> > > index 60dcfafdc11a..17210558de79 100644
> > > --- a/fs/userfaultfd.c
> > > +++ b/fs/userfaultfd.c
> > > @@ -2073,6 +2073,11 @@ static int userfaultfd_api(struct userfaultfd_ctx *ctx,
> > >   	uffdio_api.features &= ~UFFD_FEATURE_WP_UNPOPULATED;
> > >   	uffdio_api.features &= ~UFFD_FEATURE_WP_ASYNC;
> > >   #endif
> > > +
> > > +	ret = -EINVAL;
> > > +	if (features & ~uffdio_api.features)
> > > +		goto err_out;
> > > +
> > >   	uffdio_api.ioctls = UFFD_API_IOCTLS;
> > >   	ret = -EFAULT;
> > >   	if (copy_to_user(buf, &uffdio_api, sizeof(uffdio_api)))
> > 
> > CCing Peter.
> > 
> > -- 
> > Cheers,
> > 
> > David / dhildenb
> > 
> 

-- 
Peter Xu