LSI 9211-8I IT mode stuck during loading

[swagler]

vanfawx- My Motherboard only has non-AHCI SATA II and no updated drivers supporting TRIM. Which using MS Standard IDE Drivers down grades SATA to IDE-150 (but does TRIM). I went with LSI 9211-4i to draw from multiple lanes on my nForce4 PCIe v1.1 x4 slot. All PCIe AHCI SATA controllers only utilize one lane per port, with the exception of OWC Mercury Accelsior PCIe v2.0 4x SSD Adapter, which attached to v1.1 would again only draw from one lane.

 

[Terry Kennedy]

Apparently we need the SSD supports "deterministic read after trim / read zero after trim" ... I bought a 840 Pro 256GB .... and TRIM is OK !

I've been out of the disk firmware design business for years, but I'm having a hard time imagining why any SSD would not be deterministic - the controller (on the drive) knows what blocks are mapped and what ones aren't, and it should return 0's for any umapped block. Simply sending a sector full of 0's is likely faster than reading [useless] data from the NAND flash, anyway. NAND flash's erased state is normally all 1's, but there's no reason that the drive controller needs to pass that through to the user, since it is erased. Since any unmapped blocks are "fair game" for a background erase, saying "well, the controller may not know the mapping state" makes no sense as a justification, particularly since the mapping of flash blocks to "disk" blocks varies due to wear leveling.

 

[swagler]

Terry Kennedy- Thanks for input. Yes, there's an ongoing debate whether drivers need to send TRIM on newer SSD's. All I know is "deterministic read after trim / read zero after trim" is required for LSI IT Mode HBA to TRIM, Samsung 840 Pro has it, 850 onward doesn't. I just joined Lime Tech forum, there's a lot of Linux users over there discussing server configurations with LSI HBA's. I'm awaiting a reply from a user that seems knowledgeable on this issue. Below is an excerpt, apparently fstrim is Linux's built-in trimcheck.

Run fstrim -v and if you get:
FITRIM ioctl failed: Remote I/O error -> not working
29.8 GiB (32001204224 bytes) trimmed -> working

I have an 840 Pro 512GB, C: 48GB, D: 381GB, Provision Area: 47GB.
Currently C: 56% used space, D: 52% used space.

 

[vanfawx]

@swagler - Not sure if you've seen or tried this, but for fun, wanna try sg_unmap instead of fstrim? The link from Broadcom is here

 

[swagler]

Terry Kennedy- Thanks. Do you know what wdcfg.exe in Win7 driver pkg is able to configure? It looks like sg3_utils package is for Linux? I'm running Win7 Ult x64.

 

[vanfawx]

Yeah, it's Linux. Sorry, I got my wires crossed and thought you were on Linux, not Windows.

 

[swagler]

Excerpt from wikipedia:
Windows 7 initially supported TRIM only for drives in the AT Attachment family including Parallel ATA and Serial ATA, and did not support this command for any other devices including Storport PCI-Express SSDs even if the device itself would accept the command.[38] It is confirmed that with native Microsoft drivers the TRIM command works on Windows 7 in AHCI and legacy IDE / ATA Mode.[39] Windows 8 and later Windows operating systems support trim for PCI Express SSDs based on NVMe, and the unmap command which is a full analog of the TRIM command from Serial ATA for devices that use the SCSI driver stack. Microsoft has released an update for Windows 7, KB2990941, which when integrated into Windows 7 Setup using DISM, adds NVM Express support including TRIM for PCIe SSDs.

Windows 7, KB2990941 needs investigating. I see many server configurations are on Windows 2012, LSI Tech tested an older Asus board w/Windows 2012 and had no shutdown issue (CPU, fans, remain running) as I have on A8N-SLI Premium. I don't believe I tested trimcheck.exe on Win7 built-in LSI drivers. "adds NVM Express support including TRIM for PCIe SSDs" = native win7 driver TRIM support?

 

[swagler]

Confirmation: KB2990941 is only update for PCIe NVMe SSD's to have native MS driver support, it's not update for PCIe SAS/SATA connected SSD's. It sounds like Windows 8-10 support PCIe NVMe SSD's with TRIM.

 

[swagler]

I got confirmation from LSI Tech's that trimcheck.exe returns neither 9211 or 9207 controllers with v19 or v20 firmware & drivers were trimming SSD's on any Windows OS. So I removed controller, the taxing on old system wasn't worth the benefits to just drive HDD's. The performance was hardly noticed on SSD OS, but greatly improved HDD data drives. Maybe someone will benefit by my experience.

 

[swagler]

Stubborn as I am, I continued researching. Found that SAS Controllers are designed to use UNMAP, TRIM being SATA counterpart. I've found no documentation on how to configure SAS2004 to send UNMAP commands, only glossary documentation that a filter driver needs to be utilized. Likewise, I've never found just what the wdcfg.exe utility is capable of configuring i.e. UNMAP, 512 vs 4096 block config... anything of usefulness.

I found a utility called PerfectStorage 3.0, that scans every block sending UNMAP commands freeing drive space. The utility works wonderful. I'll note it only operates on mounted drives, so if you use Partition Master to resize partitions you'll need to attach SSD to motherboard so TRIM can be utilized during sizing. Samsung Magician Optimizer TRIM's entire drive, including unpartitioned area... it's recommend running periodically especially after major system changes.

TRIM and stability were main reasons I removed 9211-4i from system. I have since precisely manually configured motherboard memory timings, ran MemoryTest86, and slew of other tests, confirming system is utilizing all 4gb of RAM and preforming excellent. Support for A8N-SLI Premium was abandoned years ago, myths that A8N-SLI Premium can't utilize 4gb RAM and or is unstable with Windows 7 x64 is not true.

Only glitch remaining is 9211-4i preventing full power-off at shutdown, thus preventing auto restart as well. I run PerfectStorage daily, and after every major change/deletion of data, configured partitions so 50~% free space and 10+% over-provisioning. Clean as a cat's ass.