вт, 12 мар. 2024 г. в 21:14, Cryptearth <cryptearth@xxxxxxxxxxxxxx>: > > Hello there, > > I've recently seen a topic on the arch forums about a user reporting > some drives went missing on thier ASM1166 HBA. A reply figured out: > It's patch to limit the ports. > As I use a rather similar card with an ASM1064 I guessed: Well, could > this also affect me? And as I've just looked up: Yes, 6.8 will affect > me. How? Where are boot logs with working/not working conditions? And see official documentation on ASMedia site https://www.asmedia.com.tw/product/A58yQC9Sp5qg6TrF/58dYQ8bxZ4UR9wG5 [....] SATA interface AHCI SPEC Rev. 1.4 Four SATA Gen3 (6GBps) ports Support NCQ Support SATA LED Supported port multiplier command based switching [....] Four ports. Nothing more, nothing less. > Why? Because my card not just has one ASM1064 but also 4x JMicron > JMB575 s-ata port-multipliers which fan out the 4 root ports provided > by the ASM1064 to 20 actual usable ports, which I currently have 8 > drives connected to. Yes, if it connects each of its ports to four or five JMB575 into PMP mode - it's able to provide 16/20 SATA ports. And this patch changed nothing there. > As the user on the arch forums reports: Using the current arch kernel > with this "patch" makes the drives disappear - using older kernels it > works. I expect the same: Whenever this "patch" hits the arch standard > kernel I'll find myself with ZFS failing to import my pool - breaking > most of my system. > > In contrast to what's noted in the bug report, that the ASM chips > wrongly report more ports than they offer, or that the "patch" to just > limit the number of usable ports would "fix" anything actual does > break a lot of ASM chips - at least the two currently in the new > ahci.c file. > > As someone affected by this (I haven't tried any 6.8-rc yet - but just > from pure logic adding 1 and 1 still comes out to 2 - I can already > see what will happen) and looking at the sales numbers how popular > HBAs which uses port-multipliers are as they're quite cheap compared > to SAS controllers I highly request this "patch" to be reverted - or > at least to be gated behind a kernel flag to be able to set at boot to > disengage this port number enforcement when someone like me does use > such a weird chinese port-multiplier card. > > I don't want to argue about thier drawbacks and issues and risks - but > when comparing options getting 20 standard sata3 ports on a pci-e 3.0 > x1 card for less than 100 bucks shipped is the way to go when the > other end of the spectrum are SAS HBAs require at least a x4 slot and > special sff-cables and often come with RAID firmware not suitable for > ZFS - and yet cost at least 300 bucks USED for only 8 drives. > > Please don't force users like me to compile thier own kernel or stay > back at older versions just because someone reported some feature as a > bug - which is very likely not the ASM design itself but rather some > generic firmware shipped along with the capability to support > port-multipliers. > > Again, if you want to keep it at least add an option to disable it. > > Thank you for reading. > > Greetings from Germany > > Matt
