[SPDK] hotplugging possible bug

[SPDK] hotplugging possible bug

[William Del Genio]

[-- Attachment #1: Type: text/plain, Size: 670 bytes --]

Hi,

 

I’ve been working on hotplug events in an application using spdk and have noticed that I’ll get an error when unplugging a lot of devices in a short time, either physically or by detaching the uio_pci_generic driver:

 

nvme_uevent.c: 168:spdk_get_uevent: ***ERROR*** Socket read error(105): No buffer space available

 

Is there any issue with the uevent code?

 

Also I’m running into some occasional kernel issues ‘BUG: unable to handle kernel paging request at …’ when reattaching drives occasionally.  I’ve only seen it once using the hotplug example program, but many times using my program.

 

-William Del Genio


[-- Attachment #2: attachment.html --]
[-- Type: text/html, Size: 2899 bytes --]

Re: [SPDK] hotplugging possible bug

[Verkamp, Daniel]

[-- Attachment #1: Type: text/plain, Size: 1323 bytes --]

Errno 105 is ENOBUFS, which looks like it can occur with uevent when there is not enough space in the socket buffer; can you try increasing the ‘size’ variable in spdk_uevent_connect() to a larger value and see if the issue is still reproducible?

A “BUG: unable to handle kernel paging request” message would indicate some kind of kernel bug; do you see a backtrace in dmesg to indicate what code is causing the crash?

Thanks,
-- Daniel

From: SPDK [mailto:spdk-bounces(a)lists.01.org] On Behalf Of William Del Genio
Sent: Thursday, March 2, 2017 4:17 PM
To: Storage Performance Development Kit <spdk(a)lists.01.org>
Subject: [SPDK] hotplugging possible bug

Hi,

I’ve been working on hotplug events in an application using spdk and have noticed that I’ll get an error when unplugging a lot of devices in a short time, either physically or by detaching the uio_pci_generic driver:

nvme_uevent.c: 168:spdk_get_uevent: ***ERROR*** Socket read error(105): No buffer space available

Is there any issue with the uevent code?

Also I’m running into some occasional kernel issues ‘BUG: unable to handle kernel paging request at …’ when reattaching drives occasionally.  I’ve only seen it once using the hotplug example program, but many times using my program.

-William Del Genio

[-- Attachment #2: attachment.html --]
[-- Type: text/html, Size: 4958 bytes --]