[KERNEL][M] blu_sp★rk

[qu4nd]

Neues Buil verfügbar: blu_sp★rk r66

New build is up, wicked!

fastlane for blu_active is here with freq step selection method:
0: default conservative operation aka smooth mode
1: fast operation mode aka damn smooth mode

blu_plug also updated with up_threshold value to be a bit more lazy and match hi_speed load.

Also this build features nice updates from CAF and Nexus March updates.

Spoiler: changelog

• blu_plug: update plug threshold to match hi_speed freq load threshold
• blu_active: add fastlane freq calc selection, upstream updates
• KEYS: Fix race between read and revoke, Fix crash when attempt to garbage collect an uninstantiated keyring, Fix race between key destruction and finding a keyring by name, Permit in-place link replacement in keyring list, Perform RCU synchronisation on keys prior to key destruction, Fix handling of stored error in a negatively instantiated user key
• msm: mdss: Replace msleep() with usleep()
• vm_is_stack: use for_each_thread() rather then buggy while_each_thread()
• soc: qcom: rq_stats: Always show idle time left
• usb: host: xhci: Enable XHCI_SPURIOUS_SUCCESS for all controllers with xhci 1.0
• msm: camera: isp: protect driver from using invalid bufq
• gadget: composite: Fix crash seen when SS descriptor is not available
• msm: null pointer dereferencing
• net: wireless: bcmdhd: check packet length for event messages
• msm: arm: krait: Patch for krait array access out of bound
• net: wireless: bcmdhd: Add checks for stack buffer overflows
• bcmdhd: Add checks for stack buffer overflows

 

[qu4nd]

Es gibt mal wieder was Neues: blu_sp★rk r69

• build: use newest blu_naro-5.3, build from developer snapshot Linaro GCC 5.3-2016.03, with updated components and improvements/fixes
• f2fs: catch up with linux-3.4.y
• pipe: iovec: Fix memory corruption when retrying atomic copy as non-atomic (Android Security Advisory—2016-03-18 for CVE-2015-1805)

 

[qu4nd]

blu_sp★rk r72 ist verfügbar!

• pipe: Fix buffer offset after partially failed read (Second part of Android Security Advisory—2016-03-18 for CVE-2015-1805 patch)
• sched: Fix load avg vs cpu-hotplug
• mmc: move to a SCHED_FIFO thread
• Revert "f2fs: catch up with linux-3.4.y"

 

[bl1zz87]

Achtung v69 crashte meine ganze data Partition. Ergebnis war ein vollständiger Datenverlust da die Partitionen verschlüsselt waren. f2fs partitioniert. Nach v72 verschwanden wieder Apps sowie einzelne Daten. Ich empfehle vorerst bei Version 66 zu bleiben.

 

[schurgan]

Ich glaube bei dir ist etwas schief gelaufen, eher Einzelfall. Bei mir lief sowie mit v69 als auch mit v72 alles glatt.

 

[qu4nd]

Ich habe auch bei XDA nichts von allgemeinen Problemen dieser Art gelesen.

 

[bl1zz87]

Das kann sein. Ich wollte euch nur warnen. Bei mir trat dies auch erst nach ein paar std/Tage auf, nachdem ich mehrere Daten kopiert und verschoben habe. Bin jetzt auf 66 und seit dem kein Problem. @qu4nd ich dachte ich habe im thread bei xda sowas überflogen. Seit ihr beiden mit der Cache und data Partition auf f2fs oder ext4.

 

[qu4nd]

@bl1zz87 Ich war gedanklich in einem anderen Thread gelandet, sorry.
Habe jetzt mal in entsprechendem Thread bei XDA geschaut. Und du hast Recht! Dort berichten in der Tat Leute von Problemen mit f2fs in den aktuellen Versionen. Scheint aber auch nur für f2fs der Fall zu sein.

 

[schurgan]

Oh...das weis ich nicht. Hab mir noch nie Gedanken drüber gemacht

 

[qu4nd]

Neuer Build inkl. den Änderungen des April Updates verfügbar blu_sp★rk r73

• msm: perf: validate input argument of ev_constraints functions
• msm: vidc: Skip session id validation for HFI_EVENT_SYS_ERROR, Validate session_id in video hardware response, Make buffer validity checks stronger, Do sanity check for different IOCTL cmd
• net: wireless: bcmdhd: Do not print out device name on invalid length

 

[qu4nd]

blu_sp★rk r75

- build: use newest blu_naro-5.3, build from developer snapshot Linaro GCC 5.3-2016.04, with updated components and improvements/fixes
- fuse: break infinite loop in
- fuse_fill_write_pages()
- tty: Fix unsafe ldisc reference via ioctl(TIOCGETD)
- dm-crypt: run in a WQ_HIGHPRI workqueue
- dm-verity: run in a WQ_HIGHPRI workqueues
- small RAMdisk tweaks

 

[Bl4ck3urn]

blu_sp★rk r78

• blu_plug: Fix direct references to HZ
• Fix CONFIG_HZ dependency in wifi driver.
• deconfig: enable 300HZ
• onyx: Set CONFIG_HZ to 300
• msm: kgsl: Fix direct references to HZ
• sched_clock: Backport high-resolution support
• clocksource: Extract max nsec calculation into separate function
• jiffies: Avoid undefined behavior from signed overflow
• lib: align source before using optimized implementation
• memcg: Allow non-root users permission to control memory
• slub: fix a possible memory leak
• mm: vmscan: fix inappropriate zone congestion clearing
• mm: fix anon vma naming
• makefile: disable some warnings globaly
• ARM: mm: remove redundant dsb() prior to range TLB invalidation
• ARM: 7746/1: mm: lazy cache flushing on non-mapped pages
• ARM: 7730/1: DMA-mapping: mark all !DMA_TO_DEVICE pages in unmapping as clean
• msm_rmnet: Kill logspam
• bluetooth: Validate socket address length in sco_sock_bind().
• genirq: call cancel_work_sync from irq_set_affinity_notifier

 

[bl1zz87]

Blu spark r82 ist online!

* dt2w: fix HZ dependency
* clocksource: Fix abs() usage w/ 64bit values
* genirq: Fix race in register_irq_proc()
* sched/core: Fix TASK_DEAD race in finish_task_switch()
* module: Fix locking in symbol_put_addr()
* ARM: fix Thumb2 signal handling when ARMv6 is enabled
* ARM: 7880/1: Clear the IT state independent of the Thumb-2 mode
* ARM: 8429/1: disable GCC SRA optimization

 

[qu4nd]

blu_sp★rk r86

New build is up, enjoy!
If you're using f2fs, I recommend backup your data to external source in case you have any hiccups, it's a long update on the driver.

Quote:

• f2fs: catch up with 4.6-rc4-3.4
• msm: kgsl: Fix race condition in adreno_spin_idle()
• pipe: limit the per-user amount of pages allocated in pipes
• ASoC: msm: qdsp6v2: add integer overflow check
• defconfig: Remove MSM_BUSPM_DEV
• platform: msm: msm_bus: remove buspm support
• firmware_class: Fix firmware_loading_timeout usage of jiffies
• arm: arch_timer: Update for 64bit sched_clock
• msm: kgsl: Fix idle timer sysfs after removing references to HZ

 

[qu4nd]

blu_sp★rk r87

Revert derps from f2fs: catch up with 4.6-rc4-3.4

 

[qu4nd]

blu_sp★rk r88

I tagged it beta because GCC6 is a big jump, but we found nothing derped so far.

• kernel: add support for gcc 6
• makefile: disable some gcc6 warnings globaly

 

[qu4nd]

blu_sp★rk r91

Wegen einem Backport zu f2fs 4.7 wird f2fs Nutzern wird ein Backup der Daten nach extern empfohlen ("just in case").
Juni Security Patch ist auch drin.

• f2fs: catch up with linux-3.4.y, catch up with 4.6-rc4-3.4, catch up with 4.7-rc1-3.4
• msm: kgsl: Add missing checks for alloc size and sglen
• net: wireless: bcmdhd: Verify SSID length, Verify SSID length (part deux), check privilege on priv cmd
• ASoC: msm: audio-effects: fix stack overread and heap overwrite
• msm: mdss: fix possible out-of-bounds and overflow issue in mdp debugfs
• msm: camera: ispif: Validate VFE num input during reset
• ASoC: msm: q6: check upper bounds when copy ac3 params
• msm_fb: display: Enable display debugging through mdp debugfs
• msm: display: Validate MDP, MDDI and HDMI debug reg offset
• slub: fix incorrect return type of get_any_partial()
• USB: usbfs: fix potential infoleak in devio
• sysrq: Emergency Remount R/O in reverse order

 

[druid]

Der r91 läuft nicht rund bei mir (Reboots). Bin zurück zum r88, welcher ohne derlei Probleme läuft.

 

[noatsch]

Keine Probleme mit dem r91

 

[qu4nd]

blu_sp★rk r92

• build: use newest and improved blu_naro-6.1, build from developer snapshot Linaro GCC 6.1-2016.06, with updated components and improvements/fixes
• USB: cdc-acm: more sanity checking
• usbnet: cleanup after bind() in probe()
• ipv4: Don't do expensive useless work during inetdev destroy
• netfilter: x_tables: fix unconditional helper, check for size overflow
• arm: irq: Notify affinity change when migrating IRQs during hotplug, warn only when affinity really breaks