The upstream refactoring of ndaregister() to split out ndasetgeom() accidentally used an uninitialed variable to decide whether or not to set DISKFLAG_UNMAPPED_BIO. Fix this by moving that portion of ndasetgeom() back up to ndaregister(). The check for PIM_UNMAPPED is not really needed because nvme devices always have that set, so it cannot change in the other path that ndasetgeom() is now called. Approved by: re (cperciva) Reviewed by: cperciva, gallatin, imp Fixes: dffd882d12d2a71aca464f48209ec9ae6f393b15 Sponsored by: Netflix MFC After: 1 minute (cherry picked from commit 2b4dbad2db5766294ee97bb96228ec6826a9e7c3) (cherry picked from commit e271f9327f46250c9043c29c86e943d53080bf2a)
Register for AC_GETDEV_CHANGED. When we receive a namespace notification, we only create a new device if it was unconfigured. If it was configured, generate this async event. Rely on the fact that we reconstruct namespace to just get the data from the identify data and call disk_resised. Approved by: re (cperciva) Reviewed by: imp Differential Revision: https://reviews.freebsd.org/D33032 (cherry picked from commit 86d3ec359a56d1b5d015718bd19ef4bda681a032) (cherry picked from commit 9a465b37ea17642d45597d4ee7d3283b02dfa6f0)
Refactor setting of geometry for the disk to its own function. No functional changes. Approved by: re (cperciva) Reviewed by: imp Differential Revision: https://reviews.freebsd.org/D33032 (cherry picked from commit dffd882d12d2a71aca464f48209ec9ae6f393b15) (cherry picked from commit 31412fda9fa71d85a19604151d662ac65c301c2a)
The descriptions for these unmapped_io and rotating sysctls indicated that they're deprecated and being removed for FreeBSD 15.0. That did not happen, so update to FreeBSD 16 instead. Approved by: re (cperciva) Sponsored by: The FreeBSD Foundation (cherry picked from commit e93db9abc9a62d662c40d783663d64cdb829a0cc) (cherry picked from commit 469ab88d107c05ab533a15d4014d1a97b5a13c86)
- s/tranferred/transferred/ MFC after: 3 days
All queued CCBs should be created with a real priority (one that's not CAM_PRIORITY_NONE). Recently, I introduced a bug that revealed a latent MMC bug where it would stop enumerating due to a bad priority. Add an assert to catch that (the other bug in mmc_da that it found has been fixed). Sponsored by: Netflix
Queued CCBs usually are queued at CAM_PRIORITY_NORMAL, unless they are doing error recovery, or device enumeration of some kind. They should never be queued at CAM_PRIORITY_NONE, which should only be used for CCBs that are immediate. For sdda_start_init_task(), we allocate a ccb, initialize it then use it to talk to the SD/MMC card to query it, negotiate the speed and lane sizes, etc. Most of these commands are queued, so use the normal priority. Sponsored by: Netflix
Turns out, we don't use the results of xpt_path_inq here at all. And it also causes problems. Since it calls xpt_cam_inq to do this useless XPT_PATH_INQ, it loses the original priority we had for the CCB. This priority should be CAM_PRIORITY_XPT, but was oringially set to CAM_PRIORITY_NORMAL. This worked to enumerate the device because no normal priority CCBs were queued by anything doing the enumeration. However, when I changed xpt_path_inq to use the more proper PRIORITY_NONE, it exposed this bug because queued CCBs with PRIORITY_NONE sometimes won't run. This caused the probe device to stop after its first operation. Removing the xpt_path_inq means we no longer step on the important fields we get from xpt_schedule, allowing probing to work correctly. Noticed by: bz@ Sponsored by: Netflix
The new xpt_gdev_type() function shouldn't have had a blank line. It was copied from xpt_path_inq(). Remove it from both. Noticed by: jhb Sponsored by: Netflix
HGST disks that are sick are returning 44/0 for START UNIT (which we ignore) and then 4/2 on READ CAPACITY. START UNIT should be enough for READ CAPACITY to succeed or UNIT ATTENTION. However, we get NOT_READ + 4/2 back. I've seen this on several models of HGST drives. Invalidate the peripheral when we detect this condition. This is likely the least bad thing we can do: It removes access to daX, but leaves passY so logs may be extracted (if awkwardly). Removing daX access removes the disk device that causes problems to geom outlined below. Although the timeout is 5s for READ_CAPACITY, we wait the full 30s for READ_CAPACITY_16. This causes us to stall booting as we start to taste as soon as we release the final hold... but the tasting means g_wait_idle() takes now takes over 5 minutes to clear since we do this for all the opens. Even using a timeout of 3s instead of 30s leads to boot times of almost 5 minutes in these cases, so there are other, downstream operations that are taking a while, so it's not just a matter of adjusting the timeout. Failing the periph early solves the bulk of this problem (the tasting related delays). What the HBA does is HBA specific and some have firmwares that are also confused by this when they enumerate or discover the drive, leading to long (but still shorter than 5 minute) delays. This patch won't solve that aspect of startup delays with sick disks. Perhaps we should fail the periph when START UNIT fails with the same codes we check in the read capacity path. I'm reluctant to do such a global change since it's in cam_periph, and there seems no good way to flag that we want this behavior. It's also a bit magical when it runs (some drive report 44/0 always, and some just report it on START UNIT, and these HGST drive fall into the latter category). Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D51218