That warning doesn't tell us very much; just that there is a pinleak of the pages. Try drm.debug=0xe dmesg to see if there are any precursors.

Created attachment 136307 [details]
bo->refcnt error details and related logs

Not a capture of that warning (sorry!), but Xorg crashed with what might be a relevant assert, so attaching that and my boot log so far in case it's related.

[  4941.443] (EE) assert_bo_retired:325 assertion 'bo->refcnt' failed

(In reply to Adric Blake from comment #2)
> Created attachment 136307 [details]
> bo->refcnt error details and related logs
> 
> Not a capture of that warning (sorry!), but Xorg crashed with what might be
> a relevant assert, so attaching that and my boot log so far in case it's
> related.
> 
> [  4941.443] (EE) assert_bo_retired:325 assertion 'bo->refcnt' failed

No, that's just the buffer cache tripping over itself. The refcnt check was there to optimistically prevent accidental use-after-free, but it also is being triggered on one path having just pulled the buffer out of the free cache.

commit 708255cbca83dbafe9154c39bd4553b0b81f09cf (HEAD, upstream/master)
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Wed Dec 20 08:46:43 2017 +0000

    sna: Drop redundant sync of an inactive snoop cache
    
    Having requested an inactive snooped buffer, we know that it is
    already flushed and do not need to flush it again. With debugging
    enabled, we hit an assert while flushing that the buffer has a refcout,
    which at this present time of being plucked from the snoop cache it does
    not.
    
    Reported-by: Adric Blake <promarbler14@gmail.com>
    References: https://bugs.freedesktop.org/show_bug.cgi?id=104341#c2
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

FYI, I haven't once seen this message again. No idea whether it was kernel or software changes that fixed it

linux-git 4.16rc5.r4.gfc6eabbbf8ef-1 (not drm-tip)
xorg-server 1.19.6+13+gd0d1a694f-1
mesa 17.3.7-1
libdrm 2.4.91-3
xf86-video-intel 1:2.99.917+814+gaa36399c-1

Do you think we can close this?

First of all. Sorry about spam.
This is mass update for our bugs. 

Sorry if you feel this annoying but with this trying to understand if bug still valid or not.
If bug investigation still in progress, please ignore this and I apologize!

If you think this is not anymore valid, please comment to the bug that can be closed.
If you haven't tested with our latest pre-upstream tree(drm-tip), can you do that also to see if issue is valid there still and if you cannot see issue there, please comment to the bug.

I have confidence that this can be closed.

I appreciate your feedback, resolving.

I've finally had it recur after about four months... unfortunately, no debug info since it was out of the blue.

Here's an unhelpful warning in case you can somehow discern a pattern.
[605072.893820] arch_pc kernel: ------------[ cut here ]------------
[605072.893824] arch_pc kernel: WARN_ON(i915_gem_object_has_pinned_pages(obj))
[605072.893951] arch_pc kernel: WARNING: CPU: 1 PID: 29874 at drivers/gpu/drm/i915/i915_gem.c:4668 __i915_gem_free_objects+0x2a4/0x2e0 [i915]
[605072.893952] arch_pc kernel: Modules linked in: ccm sctp libcrc32c crc32c_generic hid_logitech_hidpp hid_logitech_dj hid_generic usbhid hid fuse xt_tcpudp ipt_REJECT nf_reject_ipv4 xt_set iptable_filter ip_set_hash_net ip_set nfnetlink arc4 b43 joydev bcma mousedev mac80211 snd_hda_codec_idt snd_hda_codec_generic cfg80211 rng_core iTCO_wdt gpio_ich iTCO_vendor_support dell_laptop wmi_bmof rfkill dell_wmi sparse_keymap snd_hda_intel dell_smbios i915 snd_hda_codec dell_wmi_descriptor dcdbas uvcvideo snd_hda_core snd_hwdep videobuf2_vmalloc dell_smm_hwmon videobuf2_memops snd_pcm coretemp snd_timer videobuf2_v4l2 ssb snd psmouse ums_realtek videobuf2_common i2c_algo_bit pcspkr uas videodev media soundcore drm_kms_helper input_leds led_class drm intel_agp mac_hid intel_gtt agpgart syscopyarea mmc_core sysfillrect i2c_i801
[605072.894013] arch_pc kernel:  pcmcia sysimgblt fb_sys_fops pcmcia_core lpc_ich shpchp wmi battery ac acpi_cpufreq evdev rtc_cmos sg crypto_user ip_tables x_tables ext4 crc16 mbcache jbd2 fscrypto sr_mod cdrom sd_mod usb_storage serio_raw atkbd libps2 uhci_hcd ahci libahci libata ehci_pci ehci_hcd scsi_mod usbcore usb_common i8042 serio sky2
[605072.894051] arch_pc kernel: CPU: 1 PID: 29874 Comm: kworker/1:0 Not tainted 4.16.2-1-ARCH #1
[605072.894052] arch_pc kernel: Hardware name: Dell Inc. Inspiron 1545                   /0G848F, BIOS A07 05/13/2009
[605072.894090] arch_pc kernel: Workqueue: events __i915_gem_free_work [i915]
[605072.894123] arch_pc kernel: RIP: 0010:__i915_gem_free_objects+0x2a4/0x2e0 [i915]
[605072.894124] arch_pc kernel: RSP: 0018:ffffb2568a97fe30 EFLAGS: 00010282
[605072.894127] arch_pc kernel: RAX: 0000000000000000 RBX: ffff9e1581fee700 RCX: 0000000000000001
[605072.894128] arch_pc kernel: RDX: 0000000080000001 RSI: 0000000000000086 RDI: 00000000ffffffff
[605072.894130] arch_pc kernel: RBP: ffff9e1581fee7f8 R08: 0000544b132ad098 R09: 00000000000004e8
[605072.894131] arch_pc kernel: R10: fffff22bc103b2c0 R11: 0000000000000000 R12: ffff9e1698ee8000
[605072.894133] arch_pc kernel: R13: fffffffffffffed8 R14: ffff9e1581fee5f8 R15: ffff9e1698ee8070
[605072.894135] arch_pc kernel: FS:  0000000000000000(0000) GS:ffff9e169fd00000(0000) knlGS:0000000000000000
[605072.894136] arch_pc kernel: CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[605072.894138] arch_pc kernel: CR2: 00007f6205085000 CR3: 00000000c448c000 CR4: 00000000000406e0
[605072.894139] arch_pc kernel: Call Trace:
[605072.894178] arch_pc kernel:  __i915_gem_free_work+0x62/0x90 [i915]
[605072.894184] arch_pc kernel:  process_one_work+0x1ce/0x3f0
[605072.894188] arch_pc kernel:  worker_thread+0x2b/0x3d0
[605072.894191] arch_pc kernel:  ? process_one_work+0x3f0/0x3f0
[605072.894193] arch_pc kernel:  kthread+0x113/0x130
[605072.894196] arch_pc kernel:  ? kthread_create_on_node+0x70/0x70
[605072.894199] arch_pc kernel:  ? do_syscall_64+0x74/0x190
[605072.894202] arch_pc kernel:  ? SyS_exit_group+0x10/0x10
[605072.894206] arch_pc kernel:  ret_from_fork+0x35/0x40
[605072.894209] arch_pc kernel: Code: e6 fd ff ff e8 55 d5 98 ed e9 dc fd ff ff e8 d4 fd ec ed e9 e1 fe ff ff 48 c7 c6 10 ce 75 c0 48 c7 c7 49 db 74 c0 e8 ac 6e a0 ed <0f> 0b c7 83 c8 01 00 00 00 00 00 00 e9 84 fe ff ff 48 8b 45 00 
[605072.894255] arch_pc kernel: ---[ end trace 97e4fd59c33fcb3b ]---

Maybe it's caused by slowness? I believe I was doing heavy I/O at the time, but I can't be certain.

Linux arch_pc 4.16.2-1-ARCH #1 SMP PREEMPT Thu Apr 12 13:51:26 UTC 2018 x86_64 GNU/Linux
xorg-server 1.19.6+13+gd0d1a694f-2
mesa 18.0.0-4 (upgraded to 18.0.1-1 about an hour prior but hadn't restarted.)
libdrm 2.4.91-3
xf86-video-intel 1:2.99.917+827+g02dfb919-1

Chris, any comments here?

Reoccurred today with kernel 4.17. It seems that the warning occurs completely randomly, independent of userspace. Unfortunately I still am unsure how to debug this without bogging down my system for over a month.

Happened once again, and while scrolling an innocent page in firefox. This time, I did have considerable memory pressure and was pushing into swap. Maybe part of the process is getting interrupted by memory reclamation/maintenance? I would figure that such operations would be uninterruptible, though.

Rather than something so esoteric, it could just be a side-effect of bug 103025. Can't be sure that they're related, but looking at the kernel sources where that bug seems to occur does suggest the possibility.

Adric can you attach latest teh latest dmesg from boot with kernel parameters drm.debug=0x1e log_buf_len=4M.

Apart from seeing this warning, do you see any other issues due to this warning? Trying to understand what is the impact of this warning.

I do not see any side effects aside from the warning. And getting drm debugging info will likely take weeks, even with drm.debug constantly on. I can't say whether the bug even affects kernel 4.18.

For now, I suppose you can resolve this as working and I'll reopen if I can capture the bug.

Closing as works for me. 
Warning here implies that memory is locked due to pinning of pages.