Testing Notes¶

Before Change¶

I created a ramdisk image of stock illumos bits and used it to boot (via PXE) a system with an Insyde BMC. Once booted, I confirmed that "inserting" an ISO file into the virtual CD-ROM messed things up. In particular, diskinfo would hang forever, trying to get a devices snapshot, and through kmdb we could see that sd was trapped in a MODE SENSE:

    [3]> ::stacks -m sd
    THREAD           STATE    SOBJ                COUNT
    fffffe00f7a23c20 SLEEP    SEMA                    1
                     swtch+0x133
                     sema_p+0x1cf
                     biowait+0xa4
                     default_physio+0x2cc
                     physio+0x11
                     scsi`scsi_uscsi_handle_cmd+0x2bd
                     sd`sd_ssc_send+0x12f
                     sd`sd_send_scsi_MODE_SENSE+0x174
                     sd`sd_get_caching_mode_page+0xb4
                     sd`sd_get_write_cache_enabled+0x6d
                     sd`sd_unit_attach+0xa84
                     sd`sdattach+0x2d
                     devi_attach+0xa1
                     attach_node+0x8b
                     i_ndi_config_node+0x12c
                     i_ddi_attachchild+0x3a
                     devi_attach_node+0x5d
                     config_immediate_children+0xd0
                     ndi_busop_bus_config+0x98
                     scsa2usb`scsa2usb_scsi_bus_config+0xbc
                     scsi`scsi_hba_bus_config+0x72
                     devi_config_common+0xcf
                     mt_config_thread+0x13a
                     thread_start+0xb

I "unplugged" the virtual CD-ROM and things immediately came good after we gave up:

    Oct  5 04:02:22 unknown scsi: WARNING: /pci@1d,0/pci1022,1484@8,1/pci1022,148c@0,3/hub@2/storage@1/disk@0,0 (sd1):
    Oct  5 04:02:22 unknown         Command failed to complete...Device is gone

After Change¶

I build another ramdisk with new bits built with this change (and #13817 upon which it depends). After inserting the CD-ROM this time, diskinfo was able to complete; NB, diskinfo does not actually list CD-ROM drives, but it at least was not hung. I was able to mount the hsfs file system on the device and read the contents.

I confirmed at the scsa2usb level that the device had the expected attribute disabled as per the quirks table:

> *scsa2usb_statep::walk softstate | ::print struct scsa2usb_state scsa2usb_dip | ::devinfo -q
fffffeb1cdd65c78 usbif,class8, instance #0 (driver name: scsa2usb)

> fffffeb1cdd65c78::devinfo ! grep -A1 usb-.*-id
            name='usb-revision-id' type=int items=1
                value=00000409
            name='usb-product-id' type=int items=1
                value=000003ea
            name='usb-vendor-id' type=int items=1
                value=00000b1f

> *scsa2usb_statep::walk softstate | ::print struct scsa2usb_state scsa2usb_attrs
scsa2usb_attrs = 0xfdff
> ~fdff=x
                200
> scsa2usb_quirks::print ! tail
        q_rev = 0xffff
        q_attr = 0x4000
    },
    {
        q_vid = 0xb1f
        q_pid = 0x3ea
        q_rev = 0xffff
        q_attr = 0x200
    },
]