Netgate 1100 / 2100 - 26.07 - bus_dmamem_alloc failed to align memory properly
-
We have a customer with a small remote office. The Netgate 1100 there is not stable. 26.07. We haven't seen this issue on any other netgate device, we mainly have 1100 and 2100 devices. This one from time to time loses its WAN connection, leaving us unable to access it. Checking the logs, during reboot of the 1100 I saw this:
After over 2000 of these 'bus_dmamem_alloc failed to align memory properly' messages, it 'just' continues with it's boot:
I've never seen this before. Before I check all logs thouroughly, what would this 'bus_dmamem_alloc failed to align memory properly' mean?
[edit]
I just checked my Netgate 2100 at my home office, and that does excactly the same running 26.07. So probably it's nothing specific to that 1100, yet I still wonder what all those messages mean. -
Hmm, so after seeing that at boot it continues to boot and run normally? No effects after boot?
-
@stephenw10 No apparent issues. I have in the meanwhile factory defaulted my own 2100, and still had the same result. Made a full log and created a ticket for it. We'll see.
-
I found where that message is coming from in the FreeBSD source:
sys/arm64/arm64/busdma_bounce.c#L618-L620
The outer function is
bounce_bus_dmamem_alloc()which describes itself as "Allocate a piece of memory that can be efficiently mapped into bus device space based on the constraints lited in the dma tag. A dmamap to for use with dmamap_load is also allocated."Funny that it seems there's a typo in that function block comment (
lited→listed?)Anyway, it looks like a fix was just committed to
busdma_bounce.c, I wonder if it'd fix this issue (my guess is yes)... -
A TAC ticket or a redmine issue? Or both?
-
The fix would have to be cherry picked from main, and I'm not sure at all how Netgate does their builds… that's above my pay grade.
-
@stephenw10 said in Netgate 1100 / 2100 - 26.07 - bus_dmamem_alloc failed to align memory properly:
A TAC ticket or a redmine issue? Or both?
A TAC ticket. It has been moved into an 'internal issue 23844'. I haven't checked if that means a redmine ticket but I guess so.
-
Don't see it on redmine, and those numbers only just passed 17000 so this must be some internal TAC issue number. @rgijsen please keep this thread posted for any other curious 1100/2100 owners…
-
@rgijsen said in Netgate 1100 / 2100 - 26.07 - bus_dmamem_alloc failed to align memory properly:
internal issue 23844
Ah, yes I see it. I'll co-ordinate there.
-
Just another data point: I manage a few 2100's (with 26.07). I decided to check 'em out to see if this error was appearing in the logs - it wasn't. So there must be some special circumstances here.
-
@luckman212 said in Netgate 1100 / 2100 - 26.07 - bus_dmamem_alloc failed to align memory properly:
Just another data point: I manage a few 2100's (with 26.07). I decided to check 'em out to see if this error was appearing in the logs - it wasn't. So there must be some special circumstances here.
Same - we have a lot of small branch-offices using measly 1100 units. Some of them have these log-entries, some don't. What I find weird is that after factory-defaulting my 2100 I DO have those errors. Oh how I long for proper installation images to return rather than the installer. Just to make sure you have the exact baseline as intended.
-
Mmm, I thought it might be ufs vs zfs but I'm seeing it on ufs installs too here. Still digging...
-
Different uboot versions perhaps? That shouldn't be if they've ever been upgraded since it gets updated at that point.
-
@stephenw10 I checked all of the 2100's I have access to now (a mix of ufs and zfs as well). They all match this U-Boot version
2018.03-devel-1.2.0ROGUE2-01.00.00.02+:
-
Hmm, not that then.....
-
Last night my 2100 lost WAN connectivity. I eventually had to power cycle the device since I was locked out of both web interface and SSH. My system logs are also filled with "bus_dmamem_alloc failed to align memory properly". This is the first time I've had an issue with my Netgate.
-
@stephenw10 Just wanted to chime in. I have a couple of 2100’s as well, and they both logged thousands of those lines during last reboot (took a good while to reboot at well), so it seems it quite common
-
I just asked for an update in the case and got one very quicly:
Hello,
Unfortunately, there have not been any updates on the internal ticket yet.
Yes, it appears that all appliances are affected. However, since the issue does not impact the general operation or stability of the devices, it is not currently classified as high severity. As a result, it has not been prioritized over issues with a greater operational impact.
We understand that this is still important to you and appreciate your continued patience while our development team works through their current priorities.
Thank you,
While I agree it doesn't break anything as it looks now, I still feel it floods the logs, recycling maybe important messages when no external syslog is used. I wonder when there will be a fix.
-
Hmm, as I understand it in fact not all aarch64 devices are affected. Which makes it more interesting because identifying any differences on unaffected devices may help solve it.
But, yes, it still appears to be ugly but harmless.
-
@stephenw10 I'm experiencing the same "bus_dmamem_alloc failed to align memory properly" issue as the others on my 2100, it's affecting the stability of my Internet connection too. Please fix this ASAP! The connection drops are really frustrating.
Privacy Policy · Cookie Policy