mjhjmtu

I am trying to get the Catalyst (Crimson) fglrx driver to work under kernel 4.18 on NixOS. I get it to compile and run under 4.18, but every time a program is started that needs to use the GPU, it will log an error. I based my package on https://aur.archlinux.org/packages/catalyst-test/.

The error I am getting under Linux 4.18 is as follows (driver seems otherwise fully functional):

kernel: BUG: Bad page map in process glxgears pte:80000005e5275025 pmd:5e5274067
kernel: page:ffffe5a017949d40 count:2 mapcount:-1536 mapping:0000000000000000 index:0x0
kernel: flags: 0x200000000000000()
kernel: raw: 0200000000000000 dead000000000100 dead000000000200 0000000000000000
kernel: raw: 0000000000000000 0000000000000000 00000002fffff9ff ffffa29c5059f000
kernel: page dumped because: bad pte
kernel: page->mem_cgroup:ffffa29c5059f000
kernel: addr:00000000bee58a73 vm_flags:040400d9 anon_vma: (null) mapping:000000006eab13d1 index:40002
kernel: file:card0 fault:ip_vm_kmap_fault [fglrx] mmap:ip_firegl_mmap [fglrx] readpage: (null)
kernel: CPU: 4 PID: 1761 Comm: glxgears Tainted: G IO 4.18.17 #1-NixOS
kernel: Hardware name: System manufacturer System Product Name/P6T DELUXE V2, BIOS 1202 12/22/2010
kernel: Call Trace:
kernel: dump_stack+0x5c/0x7b
kernel: print_bad_pte+0x1da/0x2a0
kernel: unmap_page_range+0x838/0xa10
kernel: unmap_vmas+0x4c/0xa0
kernel: unmap_region+0xae/0x110
kernel: ? firegl_ioctl+0x128/0x260 [fglrx]
kernel: ? __switch_to_asm+0x40/0x70
kernel: ? __vma_rb_erase+0x1f6/0x270
kernel: do_munmap+0x26b/0x410
kernel: vm_munmap+0x5f/0xa0
kernel: __x64_sys_munmap+0x22/0x30
kernel: do_syscall_64+0x4e/0x100
kernel: entry_SYSCALL_64_after_hwframe+0x44/0xa9
kernel: RIP: 0033:0x7fa781699487
kernel: Code: c7 c0 ff ff ff ff e9 73 ff ff ff 48 8b 15 f1 09 2c 00 f7 d8 64 89 02 e9 41 ff ff ff 0f 1f 80 00 00 00 00 b8 0b 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48>
kernel: RSP: 002b:00007ffd849c27a8 EFLAGS: 00000246 ORIG_RAX: 000000000000000b
kernel: RAX: ffffffffffffffda RBX: 0000000000000008 RCX: 00007fa781699487
kernel: RDX: 00007ffd849c27bc RSI: 0000000000001000 RDI: 00007fa77c9e4000
kernel: RBP: 0000000000001000 R08: 0000000000000000 R09: 0000000040002000
kernel: R10: 0000000000000145 R11: 0000000000000246 R12: 00007fa77c9e4000
kernel: R13: 0000000040002000 R14: 0000000000000009 R15: 0000000000000048
kernel: Disabling lock debugging due to kernel taint

I also tried older kernels to make sure it was kernel related and not the way I packaged it. Here is the error under Linux 4.14 (driver seems otherwise fully functional):

kernel: BUG: Bad page map in process glxgears pte:80000005eb63e025 pmd:60fe27067
kernel: page:ffffdda2d7ad8f80 count:2 mapcount:-511 mapping: (null) index:0x0
kernel: flags: 0x200000000000000()
kernel: raw: 0200000000000000 0000000000000000 0000000000000000 00000002fffffe00
kernel: raw: dead000000000100 dead000000000200 0000000000000000 ffff9f0c5ddbb400
kernel: page dumped because: bad pte
kernel: page->mem_cgroup:ffff9f0c5ddbb400
kernel: addr:00007f88d08b7000 vm_flags:040400d9 anon_vma: (null) mapping:ffff9f0c4bba04a8 index:40002
kernel: file:card0 fault:ip_vm_kmap_fault [fglrx] mmap:ip_firegl_mmap [fglrx] readpage: (null)
kernel: CPU: 7 PID: 1726 Comm: glxgears Tainted: G IO 4.14.79 #1-NixOS
kernel: Hardware name: System manufacturer System Product Name/P6T DELUXE V2, BIOS 1202 12/22/2010
kernel: Call Trace:
kernel: dump_stack+0x5c/0x85
kernel: print_bad_pte+0x1da/0x2a0
kernel: ? firegl_trace+0x2a/0xf0 [fglrx]
kernel: unmap_page_range+0x80e/0x9e0
kernel: unmap_vmas+0x4c/0xa0
kernel: unmap_region+0x9b/0xf0
kernel: ? firegl_get_BIOSImage_size+0x40/0x40 [fglrx]
kernel: ? __vma_rb_erase+0x1f6/0x270
kernel: do_munmap+0x26b/0x410
kernel: vm_munmap+0x4f/0x80
kernel: SyS_munmap+0x1d/0x30
kernel: do_syscall_64+0x67/0x120
kernel: entry_SYSCALL_64_after_hwframe+0x3d/0xa2
kernel: RIP: 0033:0x7f88cfcd7487
kernel: RSP: 002b:00007ffdfcc80d58 EFLAGS: 00000246 ORIG_RAX: 000000000000000b
kernel: RAX: ffffffffffffffda RBX: 0000000000000008 RCX: 00007f88cfcd7487
kernel: RDX: 00007ffdfcc80d6c RSI: 0000000000001000 RDI: 00007f88d08b7000
kernel: RBP: 0000000000001000 R08: 0000000000000000 R09: 0000000040002000
kernel: R10: 0000000000000145 R11: 0000000000000246 R12: 00007f88d08b7000
kernel: R13: 0000000040002000 R14: 0000000000000008 R15: 0000000000000040
kernel: Disabling lock debugging due to kernel taint

Here is the error under Linux 4.9 (full crash, unresponsive black screen):

kernel: ------------[ cut here ]------------
kernel: WARNING: CPU: 1 PID: 1279 at ../drivers/pci/msi.c:1071 __pci_enable_msi_range+0x3a0/0x3e0
kernel: Modules linked in: tun msr fglrx(O) ext2 mbcache snd_hda_codec_analog snd_hda_codec_generic iTCO_wdt amdkfd mxm_wmi amd_iommu_v2 amdgpu ttm intel_powerclamp coretemp intel_cstate drm_kms_helper intel_uncore iptable_nat nf_nat_ipv4 nf_nat input_leds led_class evdev joydev drm mousedev mac_hid uas nf_conntrack_ipv6 i2c_i801 nf_defrag_ipv6 i2c>
kernel: xt_tcpudp ip6table_filter ip6_tables iptable_filter snd_pcm_oss snd_mixer_oss snd_pcm snd_timer snd soundcore atkbd libps2 serio cpufreq_ondemand kvm_intel kvm irqbypass ip_tables x_tables ipv6 crc_ccitt autofs4 sr_mod cdrom sd_mod hid_generic usbhid hid pata_marvell usb_storage ahci libahci rtc_cmos libata uhci_hcd ehci_pci ehci_hcd usbco>
kernel: CPU: 1 PID: 1279 Comm: X Tainted: G IO 4.9.135 #1-NixOS
kernel: Hardware name: System manufacturer System Product Name/P6T DELUXE V2, BIOS 1202 12/22/2010
kernel: ffffb9cb44287918 ffffffff8cf01641 0000000000000000 0000000000000000
kernel: ffffb9cb44287958 ffffffff8cc717e1 0000042fc0f4b615 0000000000001002
kernel: 0000000000000000 0000000000000001 ffff9f5ba309a000 ffff9f5b91560780
kernel: Call Trace:
kernel: [<ffffffff8cf01641>] dump_stack+0x63/0x82
kernel: [<ffffffff8cc717e1>] __warn+0xe1/0x100
kernel: [<ffffffff8cc718cd>] warn_slowpath_null+0x1d/0x20
kernel: [<ffffffff8cf69490>] __pci_enable_msi_range+0x3a0/0x3e0
kernel: [<ffffffff8cf694e0>] pci_enable_msi_range+0x10/0x20
kernel: [<ffffffffc0c74478>] KCL_RequestMSI+0x18/0x30 [fglrx]
kernel: [<ffffffffc0c97bdc>] IRQMGR_initialize+0x2bc/0x410 [fglrx]
kernel: [<ffffffffc0c978c7>] ? IRQMGR_alloc_context+0xb7/0x110 [fglrx]
kernel: [<ffffffffc0c96107>] ? irqmgr_wrap_initialize+0x27/0x140 [fglrx]
kernel: [<ffffffffc0c783d2>] ? KCL_DEBUG_Print_Trace+0x22/0xd0 [fglrx]
kernel: [<ffffffffc0c976c5>] ? firegl_irqmgr_init+0x55/0x100 [fglrx]
kernel: [<ffffffffc0ca8a11>] ? hal_init_gpu+0x221/0x4e0 [fglrx]
kernel: [<ffffffffc0c81e7e>] ? firegl_open+0x2de/0x310 [fglrx]
kernel: [<ffffffffc0c70157>] ? ip_firegl_open+0x17/0x20 [fglrx]
kernel: [<ffffffffc0c719c9>] ? firegl_stub_open+0x99/0x100 [fglrx]
kernel: [<ffffffff8cdf8085>] ? chrdev_open+0xb5/0x180
kernel: [<ffffffff8cdf093c>] ? do_dentry_open+0x1cc/0x2f0
kernel: [<ffffffff8cdf7fd0>] ? cdev_put.part.0+0x20/0x20
kernel: [<ffffffff8cdf1d7c>] ? vfs_open+0x4c/0x80
kernel: [<ffffffff8cdfe87b>] ? may_open+0x9b/0x100
kernel: [<ffffffff8ce02958>] ? path_openat+0x698/0x1720
kernel: [<ffffffffc0c72e76>] ? KCL_SPINLOCK_STATIC_Grab+0x16/0x20 [fglrx]
kernel: [<ffffffffc0c7a52b>] ? drm_free+0x11b/0x1a0 [fglrx]
kernel: [<ffffffff8ce04b40>] ? do_filp_open+0x80/0xd0
kernel: [<ffffffff8cdd63d9>] ? kmem_cache_alloc+0x149/0x1a0
kernel: [<ffffffff8ce12e06>] ? __alloc_fd+0x46/0x170
kernel: [<ffffffff8cdf2232>] ? do_sys_open+0x1a2/0x230
kernel: [<ffffffff8cdf22f4>] ? SyS_openat+0x14/0x20
kernel: [<ffffffff8cc03b08>] ? do_syscall_64+0x68/0xe0
kernel: [<ffffffff8d17d8ce>] ? entry_SYSCALL_64_after_swapgs+0x58/0xc6
kernel: ---[ end trace 599b517fa3e894a7 ]---
kernel: ------------[ cut here ]------------
kernel: kernel BUG at ../drivers/pci/msi.c:371!
kernel: invalid opcode: 0000 [#1] SMP
kernel: Modules linked in: tun msr fglrx(O) ext2 mbcache snd_hda_codec_analog snd_hda_codec_generic iTCO_wdt amdkfd mxm_wmi amd_iommu_v2 amdgpu ttm intel_powerclamp coretemp intel_cstate drm_kms_helper intel_uncore iptable_nat nf_nat_ipv4 nf_nat input_leds led_class evdev joydev drm mousedev mac_hid uas nf_conntrack_ipv6 i2c_i801 nf_defrag_ipv6 i2c>
kernel: xt_tcpudp ip6table_filter ip6_tables iptable_filter snd_pcm_oss snd_mixer_oss snd_pcm snd_timer snd soundcore atkbd libps2 serio cpufreq_ondemand kvm_intel kvm irqbypass ip_tables x_tables ipv6 crc_ccitt autofs4 sr_mod cdrom sd_mod hid_generic usbhid hid pata_marvell usb_storage ahci libahci rtc_cmos libata uhci_hcd ehci_pci ehci_hcd usbco>
kernel: CPU: 1 PID: 1279 Comm: X Tainted: G W IO 4.9.135 #1-NixOS
kernel: Hardware name: System manufacturer System Product Name/P6T DELUXE V2, BIOS 1202 12/22/2010
kernel: task: ffff9f5b8d103000 task.stack: ffffb9cb44284000
kernel: RIP: 0010:[<ffffffff8cf689b2>] [<ffffffff8cf689b2>] free_msi_irqs+0x182/0x1d0
kernel: RSP: 0018:ffffb9cb44287938 EFLAGS: 00010286
kernel: RAX: ffff9f5b9ad6f800 RBX: ffff9f5b99c8b000 RCX: 0000000000000022
kernel: RDX: 0000000000000000 RSI: 0000000000000022 RDI: ffff9f5ba6800138
kernel: RBP: ffffb9cb44287968 R08: ffff9f5ba6800000 R09: ffff9f5b9ad6f800
kernel: R10: 0000000000000040 R11: 0000000000000000 R12: ffff9f5ba309a298
kernel: R13: ffff9f5ba309a000 R14: 0000000000000000 R15: 0000000000000001
kernel: FS: 00007fa360c9ce40(0000) GS:ffff9f5ba7240000(0000) knlGS:0000000000000000
kernel: CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
kernel: CR2: 00007fa35bed58d0 CR3: 000000060d15e000 CR4: 0000000000000670
kernel: Stack:
kernel: ffffb9cb44287948 0000000000000001 0000000000000000 ffff9f5b8a0c9cc0
kernel: ffff9f5ba309a000 0000000000000001 ffffb9cb442879b8 ffffffff8cf693f0
kernel: ffff9f5ba309a0a0 00000001ffffffe4 0080000000000001 0000000000001002
kernel: Call Trace:
kernel: [<ffffffff8cf693f0>] __pci_enable_msi_range+0x300/0x3e0
kernel: [<ffffffff8cf694e0>] pci_enable_msi_range+0x10/0x20
kernel: [<ffffffffc0c74478>] KCL_RequestMSI+0x18/0x30 [fglrx]
kernel: [<ffffffffc0c97bdc>] IRQMGR_initialize+0x2bc/0x410 [fglrx]
kernel: [<ffffffffc0c978c7>] ? IRQMGR_alloc_context+0xb7/0x110 [fglrx]
kernel: [<ffffffffc0c96107>] ? irqmgr_wrap_initialize+0x27/0x140 [fglrx]
kernel: [<ffffffffc0c783d2>] ? KCL_DEBUG_Print_Trace+0x22/0xd0 [fglrx]
kernel: [<ffffffffc0c976c5>] ? firegl_irqmgr_init+0x55/0x100 [fglrx]
kernel: [<ffffffffc0ca8a11>] ? hal_init_gpu+0x221/0x4e0 [fglrx]
kernel: [<ffffffffc0c81e7e>] ? firegl_open+0x2de/0x310 [fglrx]
kernel: [<ffffffffc0c70157>] ? ip_firegl_open+0x17/0x20 [fglrx]
kernel: [<ffffffffc0c719c9>] ? firegl_stub_open+0x99/0x100 [fglrx]
kernel: [<ffffffff8cdf8085>] ? chrdev_open+0xb5/0x180
kernel: [<ffffffff8cdf093c>] ? do_dentry_open+0x1cc/0x2f0
kernel: [<ffffffff8cdf7fd0>] ? cdev_put.part.0+0x20/0x20
kernel: [<ffffffff8cdf1d7c>] ? vfs_open+0x4c/0x80
kernel: [<ffffffff8cdfe87b>] ? may_open+0x9b/0x100
kernel: [<ffffffff8ce02958>] ? path_openat+0x698/0x1720
kernel: [<ffffffffc0c72e76>] ? KCL_SPINLOCK_STATIC_Grab+0x16/0x20 [fglrx]
kernel: [<ffffffffc0c7a52b>] ? drm_free+0x11b/0x1a0 [fglrx]
kernel: [<ffffffff8ce04b40>] ? do_filp_open+0x80/0xd0
kernel: [<ffffffff8cdd63d9>] ? kmem_cache_alloc+0x149/0x1a0
kernel: [<ffffffff8ce12e06>] ? __alloc_fd+0x46/0x170
kernel: [<ffffffff8cdf2232>] ? do_sys_open+0x1a2/0x230
kernel: [<ffffffff8cdf22f4>] ? SyS_openat+0x14/0x20
kernel: [<ffffffff8cc03b08>] ? do_syscall_64+0x68/0xe0
kernel: [<ffffffff8d17d8ce>] ? entry_SYSCALL_64_after_swapgs+0x58/0xc6
kernel: Code: fe ff ff 45 31 f6 eb 11 41 83 c6 01 44 39 73 14 0f 86 ca fe ff ff 8b 7b 10 44 01 f7 e8 b8 cc d5 ff 48 83 b8 90 00 00 00 00 74 dd <0f> 0b 49 8d b5 a0 00 00 00 e8 30 69 d6 ff e9 cc fe ff ff 48 8b 
kernel: RIP [<ffffffff8cf689b2>] free_msi_irqs+0x182/0x1d0
kernel: RSP <ffffb9cb44287938>
kernel: ---[ end trace 599b517fa3e894a8 ]---

And under Linux 4.4 the driver is fully functional without any errors occurring.

I really want to use a more recent Linux than 4.4, 4.14 at least, but I ought to fix those bugs if I am to be comfortable with doing so. What causes bad page map from happening consistently on startup of a GPU using program? When I tried searching for answer myself, I only found the suggestion that it could have to do with overclocking, but after undoing my overclock, the error still occurs.

If it turns out to be a matter of the driver not being correctly patched, than I still have the following plans:

1. Use the older driver 15.9, based on the AUR catalyst package, rather than catalyst-test.


2. See if other distributions, like Ubuntu or Gentoo, use a different set of patches, and try them out instead.

But before trying those, I will need to know what could be the cause of these bad page map errors, because for all I know it actually is patched correctly right now (since other than the error in the log, it does function properly).