On Fri, Feb 15, 2019 at 09:22:21AM -0800, Guenter Roeck wrote:
On Fri, Feb 15, 2019 at 07:56:22AM +0100, Greg Kroah-Hartman wrote:
On Thu, Feb 14, 2019 at 01:02:26PM -0800, Guenter Roeck wrote:
On Wed, Feb 13, 2019 at 07:38:05PM +0100, Greg Kroah-Hartman wrote:
This is the start of the stable review cycle for the 4.20.9 release. There are 50 patches in this series, all will be posted as a response to this one. If anyone has any issues with these being applied, please let me know.
Responses should be made by Fri Feb 15 18:36:30 UTC 2019. Anything received after that time might be too late.
Build results: total: 159 pass: 159 fail: 0 Qemu test results: total: 343 pass: 330 fail: 13 Failed tests: sh:rts7751r2dplus_defconfig:initrd sh:rts7751r2dplus_defconfig:ata:rootfs sh:rts7751r2dplus_defconfig:mmc:rootfs sh:rts7751r2dplus_defconfig:usb:rootfs sh:rts7751r2dplus_defconfig:usb-hub:rootfs sh:rts7751r2dplus_defconfig:usb-ohci:rootfs sh:rts7751r2dplus_defconfig:usb-ehci:rootfs sh:rts7751r2dplus_defconfig:usb-xhci:rootfs sh:rts7751r2dplus_defconfig:usb-uas-ehci:rootfs sh:rts7751r2dplus_defconfig:usb-uas-xhci:rootfs sh:rts7751r2dplus_defconfig:scsi[53C810]:rootfs sh:rts7751r2dplus_defconfig:scsi[53C895A]:rootfs sh:rts7751r2dplus_defconfig:scsi[FUSION]:rootfs
This failure gave me a bit of trouble. It is similar to the failure observed earlier with v4.4.y, but changing the C compiler version did not help (I tried 8.2.0 and 5.5.0), and changing the qemu version did not help either. Bisect points to commit 31e8a058e1f ("Revert "ext4: use ext4_write_inode() when fsyncing w/o a journal") as the culprit. No idea why that would be the case, but I repeated the bisect twice with the same results, and reverting that revert indeed fixes the problem. This is weird since one of the failing tests doesn't even mount a file system but boots from initrd. Go figure. Bisect results are below.
Guenter
# bad: [f4a86d6d2a0bdead7cf98d552481367e6356ef28] Linux 4.20.9-rc1 # good: [0788acb1a3ed1589da1768ba64b1e5c76e8cb661] Linux 4.20.8 git bisect start 'HEAD' 'v4.20.8' # good: [df6033ca3341635d8f529d6cdb6d37257df6f783] MIPS: OCTEON: don't set octeon_dma_bar_type if PCI is disabled git bisect good df6033ca3341635d8f529d6cdb6d37257df6f783 # good: [da33f30a55d08758a63de90038344f956252ea8f] drm/rockchip: rgb: update SPDX license identifier git bisect good da33f30a55d08758a63de90038344f956252ea8f # good: [3ef454cd518997a8447080fb01743aaa4d4de8ae] xfrm: Make set-mark default behavior backward compatible git bisect good 3ef454cd518997a8447080fb01743aaa4d4de8ae # bad: [b442f17368f050887aae5d525b4dd0e146440f94] libceph: avoid KEEPALIVE_PENDING races in ceph_con_keepalive() git bisect bad b442f17368f050887aae5d525b4dd0e146440f94 # bad: [31e8a058e1f9aa32254f64a9643853062141895a] Revert "ext4: use ext4_write_inode() when fsyncing w/o a journal" git bisect bad 31e8a058e1f9aa32254f64a9643853062141895a # good: [0bc4dd12c3196b01d2123f95bb1c949e5eebe483] drm/i915: Try to sanitize bogus DPLL state left over by broken SNB BIOSen git bisect good 0bc4dd12c3196b01d2123f95bb1c949e5eebe483 # first bad commit: [31e8a058e1f9aa32254f64a9643853062141895a] Revert "ext4: use ext4_write_inode() when fsyncing w/o a journal"
That's really odd, given that this is only showing up in sh and not in any of the other releases. I'll blame gcc for this one :(
I suspect it may be a code alignment issue, but who knows. The problem is gone in v4.20.10, meaning the revert of "exec: load_script: don't blindly truncate shebang string" "fixed" the problem as well.
That's sad, and scary :(
Thanks for letting me know.
greg k-h