Wiki . Report Bug · Request Feature
This is an unsupported update of the original driver source for FusionIO cards. It comes with no warranty, it may cause DATA LOSS or CORRUPTION.
For the iomemory-vsl4 driver please go to the iomemory-vsl4 repo.
Please refer to the Fusion-io and OEM Part Numbers Matrix.
Driver support for FusionIO cards has been lagging behind kernel releases, effectively making these cards an expensive paperweight when running a distribution like Ubuntu / Arch / Fedora / ProxMox which all supply newer kernels than supported by the original drivers.
Releases are tagged, and should be checked out by their tag. The release tags follow Linux Kernel versions. E.g. v5.12.1 (A Fine Boi) will work on all 5.x kernels that are 5.12 and lower, but is not guaranteed to work on 5.13. v4.20.2 - Grey Dawn supports most kernels that pre-date 5.0.
| Tag | Codename |
|---|---|
| v5.12.1 | A Fine Boi |
| v5.10.0 | Fatto Catto |
| v5.6.0 | MEGACHONKER |
| v4.20.2 | Grey Dawn |
Starting with Linux kernel 5.4.0, significant changes to the kernel were made that require additional boot time kernel flags for this driver to work. These affect AMD CPUs starting with 5.4.0, and Intel CPUs after about kernel 5.8.0.
Add the following to your /etc/default/grub: For AMD systems:
amd_iommu=on iommu=pt
For Intel system:
iommu=pt
The driver is derived from the original iomemory-vsl-3.2.16, but has several fixes and gone through rigorous cleaning of redundant unused and old code. Active development is done on master, while tags are used for releases. The new releases are all tagged from master, whereas the v4.20 releases are tagged from fio-3.2.16.1732, which is a modified version of the original codebase.
The modified source fixes several known bugs in the driver:
- silent IO drops
- crashes during driver unload
- SCSI queue settings
- BIO status handling
Commits to master are not "always" write tested, just compile tested. Releases have gone through testing with Flexible I/O Tester. Testing for page_cache errors, and generic FIO checksumming on read and write and different block sizes. Non released branches are often in process of verification and under active development.
git clone https://github.com/snuf/iomemory-vsl
cd iomemory-vsl/
git checkout <release-tag>
make module
sudo insmod root/usr/src/iomemory-vsl-3.2.16/iomemory-vsl.ko
git clone https://github.com/snuf/iomemory-vsl
cd iomemory-vsl
git checkout <release-tag>
make dpkg
git clone https://github.com/snuf/iomemory-vsl
cd iomemory-vsl/
git checkout <release-tag>
make rpm
DKMS is recommended, but installation can also be done with the created packages.
Dynamic Kernel Module Support automates away the requirement of having to repackage the kernel module with every kernel and headers update that takes place on the system. This mechanism also makes sure that the driver for the new kernel actually works, or if the source needs refreshing.
git clone https://github.com/snuf/iomemory-vsl
cd iomemory-vsl/
git checkout <release-tag>
make dkms
With fio-utils installed and passing a single device through to a VM you should see someting similair to the following...:
Found 1 ioMemory device in this system
Driver version: 3.2.16 build 1731
Adapter: Single Controller Adapter
Fusion-io ioDrive2 1.205TB, Product Number:F00-001-1T20-CS-0001, SN:1241D0259, FIO SN:1241D0259
ioDrive2 Adapter Controller, PN:PA004137009
External Power: NOT connected
PCIe Bus voltage: avg 12.02V
PCIe Bus current: avg 0.77A
PCIe Bus power: avg 9.21W
PCIe Power limit threshold: 24.75W
PCIe slot available power: unavailable
Connected ioMemory modules:
fct0: Product Number:F00-001-1T20-CS-0001, SN:1241D0259
fct0 Attached
ioDrive2 Adapter Controller, Product Number:F00-001-1T20-CS-0001, SN:1241D0259
ioDrive2 Adapter Controller, PN:PA004137009
SMP(AVR) Versions: App Version: 1.0.29.0, Boot Version: 0.0.8.1
Located in slot 0 Center of ioDrive2 Adapter Controller SN:1241D0259
Powerloss protection: protected
PCI:00:04.0, Slot Number:3
Vendor:1aed, Device:2001, Sub vendor:1aed, Sub device:2001
Firmware v7.1.15, rev 110356 Public
UEFI Option ROM v3.2.6, rev 1212 Enabled
1205.00 GBytes device size
Format: v500, 2353515625 sectors of 512 bytes
PCIe slot available power: unavailable
PCIe negotiated link: 4 lanes at 5.0 Gt/sec each, 2000.00 MBytes/sec total
Internal temperature: 64.47 degC, max 64.97 degC
Internal voltage: avg 1.02V, max 1.02V
Aux voltage: avg 2.49V, max 2.49V
Reserve space status: Healthy; Reserves: 100.00%, warn at 10.00%
Active media: 100.00%
Rated PBW: 17.00 PB, 99.46% remaining
Lifetime data volumes:
Physical bytes written: 92,553,058,889,120
Physical bytes read : 82,741,942,333,288
RAM usage:
Current: 389,636,800 bytes
Peak : 389,636,800 bytes
Contained VSUs:
fioa: ID:0, UUID:0f402a7e-9948-7a4f-a2d0-7b0cb4d1ecda
fioa State: Online, Type: block device
ID:0, UUID:0f402a7e-9948-7a4f-a2d0-7b0cb4d1ecda
1205.00 GBytes device size
Format: 2353515625 sectors of 512 bytes
Where dmesg contains the actual driver version tag of the running driver
vagrant@fio:~/iomemory-vsl/root/usr/src/iomemory-vsl-3.2.16$ dmesg | grep "ioDrive driver"
[ 3.122269] <6>fioinf ioDrive driver 5.11.10-b29020f-3.2.16.1731 loading...
The support and maintenance of this driver is made possible by the people that actively contribute or contributed to its code base or by supporting the project in other ways.
| Name | Discord | |
|---|---|---|
| Demetrius Cassidy | @Tourman | Long nights of cleaning the codebase, setting up the Discord channel and guiding people through firmware upgrades |
| Vince Fleming | Donating a 1.2TB IoDrive2 for debugging problems |
Obviously all the regulars on the Discord channel, notably @bplein, @AcquaCow and @Tear.
Oh yes and the folks that were at Fusion IO for creating this product that was way ahead of its time and delivering the integration shim sources with the driver core.
Tools and resources often used to figure out what changed, and why things are not working as they are supposed to.
| Source | |
|---|---|
| Elixir | Making the Linux source code browsable and easy to compare changes over time |
| ZOL | A source of inspiration for changes in the block layers of the linux kernel that we get to deal with |
| LKML | Sometimes the first or last resort to figure out why something changed in the kernel, or where |
| The Nvidia Forum | Because they are hot to trot they encounter problems before we do, so we get to ride on their coat tails.... sometimes....though often not |
| Ghidra | The Ghidra project from the NSA that allows a look under the covers to figure out things inside non-sourcy libs |
Join us on the Discord Server in the Wiki, or create a bug report.
Installing the fio-util, fio-common, fio-preinstall and fio-sysvinit are recommended. When moving from source 2.3.11 to source 3.2.10 the firmware HAS TO BE UPDATED to 3.2.10.20150212. For the move to 3.2.15, and 3.2.16 from 3.2.10 we've not seen any issues.
Though the newer the firmware often the better. If your card is not supported with the latest firmware drop by the Discord channel and someone might be able to help you.