Comment 1 for bug 1626394

This is the first test that boots the root disk without an initrd. Indeed booting it with debug shows that the kernel detects no hard disks then (in particular not /dev/sda1), thus the boot fails like that.

With 4.4, or with 4.8 and booting with an initrd it detects the sda hard drive again:

[ 2.374065] ata1: PATA max MWDMA2 cmd 0x1f0 ctl 0x3f6 bmdma 0xc000 irq 14
[ 2.383381] ata2: PATA max MWDMA2 cmd 0x170 ctl 0x376 bmdma 0xc008 irq 15
[ 2.563557] ata2.01: NODEV after polling detection
[ 2.583452] ata1.01: NODEV after polling detection
[ 2.600806] ata1.00: ATA-7: QEMU HARDDISK, 2.5+, max UDMA/100
[ 2.617551] ata1.00: 819200 sectors, multi 16: LBA48
[ 2.632337] ata2.00: ATAPI: QEMU DVD-ROM, 2.5+, max UDMA/100
[ 2.649576] ata2.00: configured for MWDMA2
[ 2.662298] ata1.00: configured for MWDMA2
[ 2.671218] scsi 0:0:0:0: Direct-Access ATA QEMU HARDDISK 2.5+ PQ: 0 ANSI: 5

I think this is due to this config change:

$ grep CONFIG_ATA= config-4.*
config-4.4.0-9136-generic:CONFIG_ATA=y
config-4.8.0-14-generic:CONFIG_ATA=m

Which would by and large mean that it is now impossible to boot without an initrd. I don't think this is desirable, as it could actually lead to regressions on existing machines.