Pi4 64-bit raspbian kernel for testing - Focus on Pi4

[mitu]

Both kernels (4.19.x and 5.4.x) have changed how the on-board sound module is loaded (snd_bcm2835), resulting in a sightly different ALSA configuration. Use raspi-config to configure your sound output and it should work afterwards.

[mahjongg]

please post any further feedback about PiOS64 in the sticky here: viewtopic.php?f=63&t=275372

now locking this thread, so there will be just one thread for developers to follow.

[dom]

Unlocking as this thread is still useful for reporting issues with the 64-bit kernel unrelated to PiOS64.
The users here are mostly running a 32-bit buster image with just the kernel updated to 64-bit and it's
useful to note regressions when switching between 32 and 64-bit kernels. (Something not possible on PiOS64).

[mahjongg]

agreed, I was already wondering if that would be the case....

[dickon]

Possibly, like this one...

Pi 4, 2G:

arm_64bit=1
gpu_mem=256
dtoverlay=vc4-fkms-v3d

omxplayer, as root, playing H.264:

Code: Select all

pi4 login: [  352.850885] Unable to handle kernel paging request at virtual address ffffff7fb0170e40
[  352.858976] Mem abort info:
[  352.861815]   ESR = 0x96000144
[  352.864936]   EC = 0x25: DABT (current EL), IL = 32 bits
[  352.870354]   SET = 0, FnV = 0
[  352.873459]   EA = 0, S1PTW = 0
[  352.876665] Data abort info:
[  352.879606]   ISV = 0, ISS = 0x00000144
[  352.883520]   CM = 1, WnR = 1
[  352.886548] [ffffff7fb0170e40] address between user and kernel address ranges
[  352.893822] Internal error: Oops: 96000144 [#1] PREEMPT SMP
[  352.899490] Modules linked in: aes_neon_blk crypto_simd cryptd bnep rpcsec_gss_krb5 evdev v3d gpu_sched raspberrypi_hwmon vc4 cec bcm2835_v4l2(C) videobuf2_vmalloc bcm2835_isp(C) drm_kms_helper drm drm_panel_orientation_quirks w1_gpio snd_soc_core snd_compress snd_pcm_dmaengine syscopyarea sysfillrect sysimgblt fb_sys_fops uio_pdrv_genirq uio rpivid_mem snd_bcm2835(C) snd_pcm snd_timer binfmt_misc snd bcm2835_codec(C) videobuf2_dma_contig v4l2_mem2mem bcm2835_mmal_vchiq(C)  videobuf2_memops videobuf2_v4l2 videobuf2_common videodev mc brcmfmac brcmutil sha256_generic libsha256 cfg80211 8021q garp hci_uart btbcm serdev bluetooth ecdh_generic ecc rfkill vc_sm_cma(C) w1_therm wire cn ip_tables x_tables bridge stp llc ipv6 nf_defrag_ipv6
[  352.965494] CPU: 2 PID: 85 Comm: vchiq-slot/0 Tainted: G         C        5.4.44-v8+ #1320
[  352.973899] Hardware name: Raspberry Pi 4 Model B Rev 1.1 (DT)
[  352.979827] pstate: 00000005 (nzcv daif -PAN -UAO)
[  352.984704] pc : __dma_inv_area+0x28/0x58
[  352.988777] lr : arch_sync_dma_for_cpu+0x3c/0x50
[  352.993467] sp : ffffffc011c63bd0
[  352.996832] x29: ffffffc011c63bd0 x28: ffffffc010eff130
[  353.002230] x27: ffffff807b551c00 x26: ffffffc0112d3000
[  353.007628] x25: ffffffc0112d300c x24: 0000000000000002
[  353.013027] x23: 0000000000000002 x22: 000000000000021c
[  353.018424] x21: 0000000000000002 x20: 000000000000021c
[  353.023823] x19: ffffffffb0170c60 x18: 0000000000000000
[  353.029221] x17: 0000000000000000 x16: 0000000000000000
[  353.034618] x15: 0000000000000000 x14: 0000000000000000
[  353.040016] x13: 0000000000000000 x12: 0000000000000000
[  353.045414] x11: 0000000000000000 x10: 0000000000001980
[  353.050812] x9 : ffffff807b551c68 x8 : ffffff807cd15560
[  353.056210] x7 : 0000000000000001 x6 : 000000000000021c
[  353.061607] x5 : 0000000000000001 x4 : 0000000000000000
[  353.067005] x3 : 000000000000003f x2 : 0000000000000040
[  353.072403] x1 : ffffff7fb0170e40 x0 : ffffff7fb0170c60
[  353.077801] Call trace:
[  353.080284]  __dma_inv_area+0x28/0x58
[  353.084005]  dma_direct_sync_single_for_cpu+0x9c/0xa0
[  353.089139]  dma_direct_unmap_page+0xa4/0xa8
[  353.095954]  dma_direct_unmap_sg+0x58/0x80
[  353.102547]  vchiq_complete_bulk+0xb0/0x278
[  353.109247]  slot_handler_func+0xf6c/0x1580
[  353.115942]  kthread+0xfc/0x128
[  353.121581]  ret_from_fork+0x10/0x1c
[  353.127652] Code: d1000443 ea03003f 8a230021 54000040 (d50b7e21)
[  353.136299] ---[ end trace 119a545d2f9256d7 ]---

OTOH, as an unprivved user, it works fine. Not convinced that should happen, even so. Did an rpi-update at about 1730, so running whatever was current then.

[6by9]

Possibly, like this one...

Pi 4, 2G:

arm_64bit=1
gpu_mem=256
dtoverlay=vc4-fkms-v3d

omxplayer, as root, playing H.264:

Code: Select all

pi4 login: [  352.850885] Unable to handle kernel paging request at virtual address ffffff7fb0170e40
[  352.858976] Mem abort info:
[  352.861815]   ESR = 0x96000144
[  352.864936]   EC = 0x25: DABT (current EL), IL = 32 bits
[  352.870354]   SET = 0, FnV = 0
[  352.873459]   EA = 0, S1PTW = 0
[  352.876665] Data abort info:
[  352.879606]   ISV = 0, ISS = 0x00000144
[  352.883520]   CM = 1, WnR = 1
[  352.886548] [ffffff7fb0170e40] address between user and kernel address ranges
[  352.893822] Internal error: Oops: 96000144 [#1] PREEMPT SMP
[  352.899490] Modules linked in: aes_neon_blk crypto_simd cryptd bnep rpcsec_gss_krb5 evdev v3d gpu_sched raspberrypi_hwmon vc4 cec bcm2835_v4l2(C) videobuf2_vmalloc bcm2835_isp(C) drm_kms_helper drm drm_panel_orientation_quirks w1_gpio snd_soc_core snd_compress snd_pcm_dmaengine syscopyarea sysfillrect sysimgblt fb_sys_fops uio_pdrv_genirq uio rpivid_mem snd_bcm2835(C) snd_pcm snd_timer binfmt_misc snd bcm2835_codec(C) videobuf2_dma_contig v4l2_mem2mem bcm2835_mmal_vchiq(C)  videobuf2_memops videobuf2_v4l2 videobuf2_common videodev mc brcmfmac brcmutil sha256_generic libsha256 cfg80211 8021q garp hci_uart btbcm serdev bluetooth ecdh_generic ecc rfkill vc_sm_cma(C) w1_therm wire cn ip_tables x_tables bridge stp llc ipv6 nf_defrag_ipv6
[  352.965494] CPU: 2 PID: 85 Comm: vchiq-slot/0 Tainted: G         C        5.4.44-v8+ #1320
[  352.973899] Hardware name: Raspberry Pi 4 Model B Rev 1.1 (DT)
[  352.979827] pstate: 00000005 (nzcv daif -PAN -UAO)
[  352.984704] pc : __dma_inv_area+0x28/0x58
[  352.988777] lr : arch_sync_dma_for_cpu+0x3c/0x50
[  352.993467] sp : ffffffc011c63bd0
[  352.996832] x29: ffffffc011c63bd0 x28: ffffffc010eff130
[  353.002230] x27: ffffff807b551c00 x26: ffffffc0112d3000
[  353.007628] x25: ffffffc0112d300c x24: 0000000000000002
[  353.013027] x23: 0000000000000002 x22: 000000000000021c
[  353.018424] x21: 0000000000000002 x20: 000000000000021c
[  353.023823] x19: ffffffffb0170c60 x18: 0000000000000000
[  353.029221] x17: 0000000000000000 x16: 0000000000000000
[  353.034618] x15: 0000000000000000 x14: 0000000000000000
[  353.040016] x13: 0000000000000000 x12: 0000000000000000
[  353.045414] x11: 0000000000000000 x10: 0000000000001980
[  353.050812] x9 : ffffff807b551c68 x8 : ffffff807cd15560
[  353.056210] x7 : 0000000000000001 x6 : 000000000000021c
[  353.061607] x5 : 0000000000000001 x4 : 0000000000000000
[  353.067005] x3 : 000000000000003f x2 : 0000000000000040
[  353.072403] x1 : ffffff7fb0170e40 x0 : ffffff7fb0170c60
[  353.077801] Call trace:
[  353.080284]  __dma_inv_area+0x28/0x58
[  353.084005]  dma_direct_sync_single_for_cpu+0x9c/0xa0
[  353.089139]  dma_direct_unmap_page+0xa4/0xa8
[  353.095954]  dma_direct_unmap_sg+0x58/0x80
[  353.102547]  vchiq_complete_bulk+0xb0/0x278
[  353.109247]  slot_handler_func+0xf6c/0x1580
[  353.115942]  kthread+0xfc/0x128
[  353.121581]  ret_from_fork+0x10/0x1c
[  353.127652] Code: d1000443 ea03003f 8a230021 54000040 (d50b7e21)
[  353.136299] ---[ end trace 119a545d2f9256d7 ]---

OTOH, as an unprivved user, it works fine. Not convinced that should happen, even so. Did an rpi-update at about 1730, so running whatever was current then.

There was a fix to vchi yesterday for a dma_unmap_sg being done against the wrong device. It may not have made it to rpi-update as yet.

[dickon]

Ta.

It isn't an issue -- works fine as a normal user, which is a bit surprising -- but thought I'd mention it.

[jiapei100]

I'm trying to run docker on the 32-bit Raspbian with 64-bit Kernel. Unlike the other 32-bit Raspbian with 32-bit kernel, when I try:

Code: Select all

pi@pi01:~ $ docker run -it -d arm64v8/alpine
ccda6c95f81de8ac3bdb1aff69ac0a7e0059fc21f6b57206e8bb004d1e87ef5c
pi@pi01:~ $ docker ps -a
CONTAINER ID        IMAGE               COMMAND             CREATED             STATUS                       PORTS               NAMES
ccda6c95f81d        arm64v8/alpine      "/bin/sh"           9 seconds ago       Exited (159) 3 seconds ago                       nifty_gates

The running container is exited promtly, normally within 3 seconds.

This happens ONLY on this 64bit kernel Raspberry Pi4.

Did anybody have the same issue?

Cheers

[jiapei100]

It looks with this P4 64-bit Kernel, raspistill and raspivid don't work properly. Even **^C** and **^Z** won't be able to halt the camera application. Did anybody meet the same issue?

Code: Select all

pi@pi01:~ $ raspistill -o cam.jpg

Message from syslogd@pi01 at Jun 10 17:45:42 ...
 kernel:[  236.233099] Internal error: Oops: 96000144 [#1] PREEMPT SMP

Message from syslogd@pi01 at Jun 10 17:45:42 ...
 kernel:[  236.233916] Code: d1000443 ea03003f 8a230021 54000040 (d50b7e21) 
 ^Cmmal: Aborting program

^C^C^C^C^C^Z^Z^Z^Z^Z

[fanoush]

Did anybody have the same issue?

yes, search this topic for 'docker'

[fanoush]

Currently I am running the regular 4.19.97 kernel (as v8+ in 64bit mode) and everything works fine, as in 32bit mode. If I do an apt upgrade to the recently released 4.19.118 kernel video works fine, but analog audio is dead and I can't get it work again, despite numerous experiments. This applies to both 32 and 64 bit mode.

try uninstalling pulseaudio, that solved it for me

[user238424]

After

Code: Select all

sudo rpi-update

and after adding

Code: Select all

arm_64bit=1

in config.txt, and after reboot (also hard reboot), i cant connect with Putty and FTP anymore. HDMI also outputs nothing. Any fix to this?

[dom]

in config.txt, and after reboot (also hard reboot), i cant connect with Putty and FTP anymore. HDMI also outputs nothing. Any fix to this?

If you remove arm_64bit=1 (e.g. using a PC with sdcard reader) does it boot?

[dom]

It looks with this P4 64-bit Kernel, raspistill and raspivid don't work properly. Even **^C** and **^Z** won't be able to halt the camera application. Did anybody meet the same issue?

There is no mmal camera or video acceleration support currently. Only v4l2 is working with 64-bit. raspivid/raspistill uses mmal.

[user238424]

in config.txt, and after reboot (also hard reboot), i cant connect with Putty and FTP anymore. HDMI also outputs nothing. Any fix to this?

If you remove arm_64bit=1 (e.g. using a PC with sdcard reader) does it boot?

That worked, thanks a lot. By the way, that happened on Model 3b.

[thatchunkylad198966]

in config.txt, and after reboot (also hard reboot), i cant connect with Putty and FTP anymore. HDMI also outputs nothing. Any fix to this?

If you remove arm_64bit=1 (e.g. using a PC with sdcard reader) does it boot?

That worked, thanks a lot. By the way, that happened on Model 3b.

Maybe you missed the topic title:

Pi4 64-bit raspbian kernel for testing - Focus on Pi4

[dickon]

It is still a little surprising that the VC bootloader will happily attempt to boot an aarch64 kernel on its 32b-only ARMs, on older Pis. Got bitten by that one myself.

[Sgb999]

how do i go back to 32-bit? having issues with linux headers

[dom]

how do i go back to 32-bit? having issues with linux headers

Just remove "arm_64bit=1" from config.txt.

[bizzymizzyb]

i'm getting this error when trying to run gadmin-openvpn-client though I'm uncertain whether or not its due to the 64-bit kernel.

Code: Select all

Starting program: /usr/sbin/gadmin-openvpn-client 
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/arm-linux-gnueabihf/libthread_db.so.1".
[Detaching after fork from child process 27390]

Program received signal SIGSEGV, Segmentation fault.
0xf7fac1dc in strlen () from /usr/lib/arm-linux-gnueabihf/libarmmem-v8l.so

[dom]

i'm getting this error when trying to run gadmin-openvpn-client though I'm uncertain whether or not its due to the 64-bit kernel.

Are you using 32-bit raspbian usespace with 64-bit kernel enabled as described in this thread?
If so just remove arm_64bit=1 from config.txt, reboot and check if issue still happens.

[bizzymizzyb]

i'm getting this error when trying to run gadmin-openvpn-client though I'm uncertain whether or not its due to the 64-bit kernel.

Are you using 32-bit raspbian usespace with 64-bit kernel enabled as described in this thread?
If so just remove arm_64bit=1 from config.txt, reboot and check if issue still happens.

ahhhh... that's too easy - why not run GDB on it and see exactly where the error occurs? lol ; honestly I'd hate to do that:

screenshot2020062501.png (89.32 KiB) Viewed 2983 times

look at that uptime- 12 days - this system is solid as a rock. i'm not taking this baby down because of some silly, contained seg fault preventing me from running gtk config of openvpn ; the command-line version works.

But on a more serious note, there may be an issue with /usr/lib/arm-linux-gnueabihf/libarmmem-v8l.so running in 64-bit. I've never coded arm assmebly but looking at it:

Code: Select all

   0xf7fac1c0 <+0>:	push	{r4, lr}
   0xf7fac1c4 <+4>:	mov	r1, r0
   0xf7fac1c8 <+8>:	tst	r0, #248, 30	; 0x3e0
   0xf7fac1cc <+12>:	tst	r0, #7
   0xf7fac1d0 <+16>:	bne	0xf7fac24c <strlen+140>
   0xf7fac1d4 <+20>:	bics	r3, r2, r0
   0xf7fac1d8 <+24>:	beq	0xf7fac24c <strlen+140>
=> 0xf7fac1dc <+28>:	vld1.8	{d0-d1}, [r0 :64]!
   0xf7fac1e0 <+32>:	bics	r3, r2, r0
   0xf7fac1e4 <+36>:	beq	0xf7fac220 <strlen+96>
   0xf7fac1e8 <+40>:	vceq.i8	d0, d0, #0
   0xf7fac1ec <+44>:	vceq.i8	d1, d1, #0
   0xf7fac1f0 <+48>:	vmov	r3, s0
   0xf7fac1f4 <+52>:	vmov	r4, s1
   0xf7fac1f8 <+56>:	vmov	r12, s2
   0xf7fac1fc <+60>:	vmov	lr, s3
   0xf7fac200 <+64>:	teq	r3, #0
   0xf7fac204 <+68>:	teqeq	r4, #0
   0xf7fac208 <+72>:	teqeq	r12, #0
   0xf7fac20c <+76>:	teqeq	lr, #0
   0xf7fac210 <+80>:	bne	0xf7fac2d0 <strlen+272>
   0xf7fac214 <+84>:	vld1.8	{d0-d1}, [r0 :64]!

Code: Select all

r0             0x0                 0
r1             0x0                 0
r2             0xff8               4088
r3             0xff8               4088
r4             0x7fa338            8364856
r5             0x0                 0
r6             0x851d60            8723808
r7             0x798ea0            7966368
r8             0x45d100            4575488
r9             0x7f85f8            8357368
r10            0x424c50            4344912
r11            0x0                 0
r12            0x424edc            4345564
sp             0xfffef560          0xfffef560
lr             0x40d548            4248904
pc             0xf7fac1dc          0xf7fac1dc <strlen+28>
cpsr           0x20000010          536870928
d0             {u8 = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xff}, u16 = {0x0, 0x0, 0x0, 0xff00}, u32 = {0x0, 0xff000000}, u64 = 0xff00000000000000, f32 = {0x0, 0x0}, f64 = 0x8000000000000000}
d1             {u8 = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff}, u16 = {0xffff, 0xffff, 0xffff, 0xffff}, u32 = {0xffffffff, 0xffffffff}, u64 = 0xffffffffffffffff, f32 = {0x0, 0x0}, f64 = 0x0}

see anything? r0 looks suspicious but I don't know what vld1.8 does.

[dom]

libarmmem*.so is just an accelerated set of memset/memcpy style functions.
You can disable it by commenting out

Code: Select all

/usr/lib/arm-linux-gnueabihf/libarmmem-${PLATFORM}.so

from /etc/ld.so.preload and then run "sudo ldconfig"

If it still crashes (but now libc) then it's the application at fault (presumably passing in a NULL or otherwise duff pointer).
If it works then there's a bug in libarmmem.

[bizzymizzyb]

still crashes

Code: Select all

Starting program: /usr/sbin/gadmin-openvpn-client 
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/arm-linux-gnueabihf/libthread_db.so.1".
[Detaching after fork from child process 25362]

Program received signal SIGSEGV, Segmentation fault.
strlen () at ../sysdeps/arm/armv6/strlen.S:26
26	../sysdeps/arm/armv6/strlen.S: No such file or directory.

guess the problem is in gadmin-openvpn-client, same null value in r0

Code: Select all

=> 0xf745f830 <+0>:	ldrb	r2, [r0]
   0xf745f834 <+4>:	and	r3, r0, #7
   0xf745f838 <+8>:	mov	r1, r0
   0xf745f83c <+12>:	rsb	r3, r3, #15
   0xf745f840 <+16>:	cmp	r2, #0
r0             0x0                 0
r1             0x5                 5
 

surprised no one noticed it before. As I recall, the KDE openvpn config was giving a segfault as well but I didn't investigate it.

[22over7]

- Mathematica and Wolfram are not showing up

If you are daring, edit /opt/Wolfram/WolframEngine/12.0/Executables/mathematica
find lines:

Code: Select all

				armv?l)
					SystemIDList="Linux-ARM";;

and add after:

Code: Select all

				aarch64)
					SystemIDList="Linux-ARM";;

You can also make a comparable change to /opt/Wolfram/WolframEngine/12.0/Executables/wolfram

I know that this is an old post, but I wonder if we can apply this modification to the 64 bit raspberry OS to enable installing of mathematica? For my intended applications having 64 bit OS could be useful, but I need mathematica as well.

My pi has not yet arrived so I can't test this myself. Many thanks for your help!