- linaro-kernel - lists.linaro.org

Is there a mailing list for ask.linaro.org?

by Viresh Kumar

Hi guys, Sorry if i am on the wrong list :( I found it hard to log on to ask.linaro.org to see if i can answer some queries. Or even to ask some queries. I would love to have a mail + mailing-list interface for it. Please point me to one if available. -- viresh

12 years, 7 months

2
2
0 0

[PATCH V4 0/2] Upstream ST-Ericsson thermal driver

by hongbo.zhang

From: "hongbo.zhang" <hongbo.zhang(a)linaro.com> V3->V4 Changes: 1. In previous patch set V3 "Fix thermal bugs and Upstream ST-Ericsson thermal driver", there were 5 patches in total, since the first 3 for fixing thermal layer bugs have been accepted by the maintainer, I'd like to send out the updated last 2 only this time, for upstreaming ST-Ericsson thermal driver. 2. In patch "Thermal: Add ST-Ericsson DB8500 thermal driver" - typo in commit message: diver -> driver; - add more explanation of device tree properties are optional or required. - in function db8500_thermal_match_cdev: remove cdev_name, use trip_points->cdev_name[i] directly instead; - in function db8500_cdev_bind: move cdev->ops->get_max_state(cdev, &max_state) out of for loop to call it only once; - in function db8500_thermal_parse_dt: update checking strlen(tmp_str) >= THERMAL_NAME_LENGTH - add Reviewed-by: Viresh Kumar <viresh.kumar(a)linaro.org> - add Reviewed-by: Francesco Lavra <francescolavra.fl(a)gmail.com> 3. In patch "Thermal: Add ST-Ericsson DB8500 thermal properties and platform data" - add Reviewed-by: Viresh Kumar <viresh.kumar(a)linaro.org> hongbo.zhang (2): Thermal: Add ST-Ericsson DB8500 thermal driver. Thermal: Add ST-Ericsson DB8500 thermal properties and platform data. .../devicetree/bindings/thermal/db8500-thermal.txt | 44 ++ arch/arm/boot/dts/dbx5x0.dtsi | 14 + arch/arm/boot/dts/snowball.dts | 31 ++ arch/arm/configs/u8500_defconfig | 2 + arch/arm/mach-ux500/board-mop500.c | 64 +++ drivers/thermal/Kconfig | 20 + drivers/thermal/Makefile | 2 + drivers/thermal/db8500_cpufreq_cooling.c | 108 +++++ drivers/thermal/db8500_thermal.c | 530 +++++++++++++++++++++ include/linux/platform_data/db8500_thermal.h | 38 ++ 10 files changed, 853 insertions(+) create mode 100644 Documentation/devicetree/bindings/thermal/db8500-thermal.txt create mode 100644 drivers/thermal/db8500_cpufreq_cooling.c create mode 100644 drivers/thermal/db8500_thermal.c create mode 100644 include/linux/platform_data/db8500_thermal.h -- 1.7.11.3

12 years, 7 months

1
2
0 0

[ACTIVITY] (Shawn Guo) Week of Oct 29

by Shawn Guo

=== Highlights === * Sent imx multi-platform series to arm-soc for 3.8 merge window. * Sent a series adding cpuidle driver for imx6q using coupled cpuidle, implementation 3 C-states - wfi, clock gating and power gating. * Gave FSL internal development team a training on device tree, and spent time on the kernel 3.5 upgrade preparation. Shawn

12 years, 7 months

1
0
0 0

Do spinlocks really need DMB?

by Peter Fordham

Hi, I'm working with a dual Cortex-A9 system and I've been trying to understand exactly why spinlock functions need to use DMB. It seems that as long as the merging store buffer is flushed the lock value should end up in the L1 on the unlocking core and the SCU should either invalidate or update the value in the L1 of the other core. This is enough to maintain coherency and safe locking right? And doesn't STREX skip the merging store buffer anyway, meaning we don't even need the flush? DMB appears to be something of a blunt hammer, especially since it defaults to the system domain, which likely means a write all the way to main memory, which can be expensive. Are the DMBs in the locks there as a workaround for drivers that don't use smp_mb properly? I'm currently seeing, based on the performance counters, about 5% of my system cycles disappearing in stalls caused by DMB. -Pete

12 years, 7 months

3
7
0 0

[ACTIVITY] (Anton Vorontsov) 2012-10-08 - 2012-10-26

by Anton Vorontsov

== Highlights == * Got some positive reviews from Mel Gorman on the first RFC of vmevent pressure attribute. Fixed most of implementation comments; * Prepared and sent the second RFC of the pressure attribute support. The new patch integrates Minchan's and Mel's ideas of using discrete pressure levels. Also, now we use reclaimer's priority to detect pre-OOM conditions. The second RFC got some positive feedback from Minchan and Pekka; * Sent accuracy improvements for the vmevent_fd (this fixes a long standing issue noticed by Kosaki Motohiro); * vmevent_fd cleanups were merged into Pekka Enberg's tree; * All the pstore debugfs reworks landed into v3.7-rc1; * Prepared two set of slides for the upcoming Connect: status updates for my items, plus some generic notes on upstreaming; * Resent ARM KDB FIQ support patches, ARM FIQ cleanups and KDB Kiosk mode patches. Still no feedback; * Got some very nice patches for pstore: MMC backend and a tasklist dumper; briefly reviewed the patches, and hopefully they'll be merged in v3.8.

12 years, 7 months

1
0
0 0

[RFC v2 0/2] vmevent: A bit reworked pressure attribute + docs + man page

by Anton Vorontsov

Hi all, So this is the second RFC. The main change is that I decided to go with discrete levels of the pressure. When I started writing the man page, I had to describe the 'reclaimer inefficiency index', and while doing this I realized that I'm describing how the kernel is doing the memory management, which we try to avoid in the vmevent. And applications don't really care about these details: reclaimers, its inefficiency indexes, scanning window sizes, priority levels, etc. -- it's all "not interesting", and purely kernel's stuff. So I guess Mel Gorman was right, we need some sort of levels. What applications (well, activity managers) are really interested in is this: 1. Do we we sacrifice resources for new memory allocations (e.g. files cache)? 2. Does the new memory allocations' cost becomes too high, and the system hurts because of this? 3. Are we about to OOM soon? And here are the answers: 1. VMEVENT_PRESSURE_LOW 2. VMEVENT_PRESSURE_MED 3. VMEVENT_PRESSURE_OOM There is no "high" pressure, since I really don't see any definition of it, but it's possible to introduce new levels without breaking ABI. The levels described in more details in the patches, and the stuff is still tunable, but now via sysctls, not the vmevent_fd() call itself (i.e. we don't need to rebuild applications to adjust window size or other mm "details"). What I couldn't fix in this RFC is making vmevent_{scanned,reclaimed} stuff per-CPU (there's a comment describing the problem with this). But I made it lockless and tried to make it very lightweight (plus I moved the vmevent_pressure() call to a more "cold" path). Thanks, Anton.

12 years, 7 months

3
12
0 0

[ACTIVITY] (Linus Walleij) 2012-09-15 - 2012-09-23

by Linus Walleij

== Linus Walleij linusw == === Highlights === * Drilled into the RCU lockup. With some help from Paul McKenney we got a task dump from CPU0 (the CPU that is locking) but not much more - only the swapper thread is running there and we need to dump its stack. * Puffed and pinged on the SPARSE_IRQ patch, yet no reaction so far. * Mainlining some stuff in the PL011 driver that has been sitting internally for some time. Easy stuff went in first and then we ran into the trickier patches - some odd baudrates are impossible since th TTY layer will "snap" the baudrate to a standard baudrate from a table if it's "close enough". We used to work around this by not calling out to the TTY layer in the PL011 driver for certain baudrates, but this is duct-taping. Still debating the real solution with Alan Cox and Russell King who help out a lot. * Reviewed and merged various GPIO patches. * Reviewed and merged various pinctrl patches. * Reviewed some ux500 patches coming from Lee. * Reviewed Arnd's iomem cleanups. * Sent out Patrice Chotard's pinctrl patch for the PL022 SPI driver, Mark has merged it. === Plans === * Ongoing work on sparse IRQ for Nomadik and Ux500. * Test the PL08x patches on the Ericsson Research PB11MPCore and submit platform data for using pl08x DMA on that platform. * Look into other Ux500 stuff in need of mainlining... using an internal tracking sheet for this. * Look into regmap. Try something out, get to know it. === Issues === * N/A Thanks, Linus Walleij

12 years, 7 months

5
16
0 0

Connect session: "Kernel Development with Linaro CI and LAVA"

by Michael Hudson-Doyle

Hi all, We'd like to let you know about a session at Connect: http://summit.linaro.org/lce12/meeting/21231/kernel-development-with-linaro… We want to make your jobs easier -- let us know how! Cheers, mwh

12 years, 7 months

2
2
0 0

[ACTIVITY] (Linus Walleij) 2012-10-13 - 2012-10-22

by Linus Walleij

== Linus Walleij linusw == I realized that my [ACTIVITY] report appears to trigger everyone else to send theirs so when I missed mine, no reports came in, sorry! Here it goes, a weekend late, but still probably the first one. === Highlights === * SPARSE_IRQ patch set for Ux500 ready and rolling, now ready to look at multiplatform for Ux500. * Reviewed and merged some GPIO patches, wrote some GPIO driver patches trying to clean out some irqdomain business, mainly because I don't fully understand the semantics of that API myself. Now I think I do. * Wrote a few U300 patches cleaning up some pinctrl business (irqdomains, nocheinmal). * Again debated issues around runtime PM / suspend and resume proper in the different drivers. We have planned a runtime PM session @connect with Ulf and Amit. * Some heavy lifting as internal adivisor and reviewer of the pinctrl switch for the company. === Plans === * Test the PL08x patches on the Ericsson Research PB11MPCore and submit platform data for using pl08x DMA on that platform. * Look into other Ux500 stuff in need of mainlining... using an internal tracking sheet for this. * Look into regmap. Try something out, get to know it. === Issues === N/A Thanks, Linus Walleij

12 years, 7 months

1
0
0 0

linker error when using modulo operator in kernel

by Venkatraman S

After making some patch which adds a x = y % z statement in kernel, I get this linker error when trying to compile for Pandboard (omap2plus) undefined reference to `__aeabi_uldivmod' make[1]: *** [vmlinux] Error 1 make: *** [sub-make] Error 2 Simple search revealed that it's been reported before in linaro gcc and upstream gcc as well. https://bugs.launchpad.net/gcc-linaro/+bug/771832 -> still not resolved. I'm surprised that kernel doesn't use modulo for unsigned long anywhere else except in my patch. Has anyone else seen this or some weird issue in my (typical desktop) setup ? Any workarounds ? TIA, Venkat.

12 years, 7 months

1
0
0 0

Multi-port USB serial cable

by Deepak Saxena

Hi all, For the single zImage demo @ Connect, our tentative plan is to have multiple boards connected to a single host so we can just have one monitor displaying multiple console sessions and then run 'uname -a' to show that the kernels are the same (this will not win any awards for visual appeal). I don't want to spend a week trying to get a driver to work so looking for suggestions on specific brands of USB multi-port serial adapters that are known to 100% work with Linux. As in you've tried it, maybe own one (hey, how about your bring it with you?!), and it worked, not as in "I read on google that it should work" :). Thanks, ~Deepak

12 years, 7 months

4
3
0 0

[PATCH resend 0/10] Get rid of FIQ_START/enable/disable_fiq() + some FIQ cleanups

by Anton Vorontsov

Hello Russell, As the merge window is over, and hopefully there are no more serious regressions in the ARM land :-), maybe you'll find time to look into these small cleanups... The rationale is the same: During KDB FIQ patches review you mentioned that I should not introduce another FIQ_START. It seems that in v3.6-rc the FIQ_START issue was somewhat band-aided, i.e. machines don't necessary need to define this stuff any longer, but I also read the background of the issue, and you once said that you don't want the FIQ subsystem to mess with genirq. It is surely makes sense as FIQs are arch-specific, plus FIQs are special in that way that platform makers are free to connects device directly to the FIQ line, avoiding IC routing, and then enable_fiq(<IRQ>) would look awkward, at the best. So, given that, it seems that the only entity that knows for sure how the particular FIQ is routed is whoever claims the FIQ and fills-in the fiq_handler struct. It might make sense to introduce disable/enable callbacks in the fiq_handler struct, and then prototypes for disable/ enable FIQ calls would look like this: enable_fiq(struct fiq_handler *fh); disable_fiq(struct fiq_handler *fh); I.e. completely abstracted from the genirq. Although, to not duplicate IC code, I think it's pretty legitimate to call genirq functions iff the caller knows for sure that FIQ comes from IRQ line and it is is routed through the well-known platform IC. In that case (which is a majority), the typical user would do this: static int irq_line; void foo_enable_fiq(struct fiq_handler *fh) { enable_irq(irq_line); } static struct fiq_handler foo_fiq = { .name = "foo", .enable_fiq = foo_enable_fiq, }; claim_fiq(&foo_fiq); enable_fiq(&foo_fiq); Obviously, it's needless indirection, so I don't do this. If we ever pass 'foo_fiq' to some device, which does not know all the routing details, then it would make sense to introduce it, but not now. So, the patch set is pretty straightforward: - Get rid of FIQ_START. Nobody but platform-specific code (and drivers) should know the details about which interrupt can be routed to FIQ and which cannot; - Remove disable/enable_fiq() calls from the FIQ subsys (the calls can be reintroduced w/ new prototypes when/if needed). Does the approach make sense? Thanks! -- arch/arm/include/asm/fiq.h | 2 - arch/arm/include/asm/mach/irq.h | 7 +++- arch/arm/kernel/fiq.c | 37 +++++-------------- arch/arm/kernel/irq.c | 2 - arch/arm/mach-omap1/include/mach/irqs.h | 4 -- arch/arm/mach-rpc/dma.c | 4 +- arch/arm/mach-rpc/include/mach/irqs.h | 8 ++-- arch/arm/mach-rpc/irq.c | 21 ++--------- arch/arm/mach-s3c24xx/include/mach/irqs.h | 3 -- arch/arm/plat-mxc/avic.c | 5 --- arch/arm/plat-mxc/include/mach/irqs.h | 2 - arch/arm/plat-mxc/tzic.c | 5 --- arch/arm/plat-s3c24xx/irq.c | 6 +-- .../media/platform/soc_camera/mx1_camera.c | 6 +-- sound/soc/fsl/imx-pcm-fiq.c | 4 +- 15 files changed, 31 insertions(+), 85 deletions(-)

12 years, 7 months

3
13
0 0

[ACTIVITY] (Deepak Saxena) Oct 10 - Oct 13

by Deepak Saxena

== Deepak Saxena <dsaxena> == === Highlights === * Finishing up Connect session planning * Continued work on cleaning up existing BPs * Working on breaking down single zImage work into BPs * Completed https://wiki.linaro.org/WorkingGroups/Kernel/BlueprintProcess * Created new roadmap card for eMMC 4.6 PM work * Various emails and discussions on roadmap/BP process. === Plans === * Closeout on creating Android upstreaming roadmap cards * Start work on Android Upstreaming slides for Connect * Sort out Connect scheduling for KWG sessions * Usual BPs, meetings, etc... == Issues === === Travel/Time Off === * Off this Friday Oct, 19th * Connect Copenhagen

12 years, 7 months

1
0
0 0

[ACTIVITY] (Dong Aisheng) 2012-10-08 - 2012-10-13

by Dong Aisheng

=== Highlights === * KVM/ARM VGIC state saving and restoring implementation. KVM is a big new picture to me and i spent a lot time to read docs and codes to better understand it and setup test and debug environment. For this week, i mainly completed reading both vgic code and qemu in-kernel vgic implmentation code for trying to know better what exactly states need to save and restore for vgic and how save/restore works in system. Since the kvm and qemu system is big, probably i need one more week to know better on what's going on on the whole process of virtual irq handling process via vgic. I feel regret for the show process about this work. * Help Reviewed [PATCH] flexcan: disable bus error interrupts for the i.MX28. from Wolfgang Grandegger. The issue was caused by an abnormal behavior of IP that it can not generate rx/tx warning interrupt, finally we found it's IC HW issue with our IP module owner's help that for mx25/35/53 the rx/tx interrupt line actually are not connected to ARM. * Sent a simple regmap core fix patch: [PATCH 1/1] regmap: select REGMAP if REGMAP_MMIO and REGMAP_IRQ enabled

12 years, 7 months

1
0
0 0

Kernel storage team weekly meeting minutes (Oct 12, 2012)

by Venkatraman S

Attendees: Alex Lemberg, Arnd Bergmann, Balaji TK, Jakub Pavelek, Luca Porzio, Venkatraman S (#linaro-storage: alex, arnd, balajitk, Xruxa, lupo, svenkatr) Discussion Notes:- 1. Mini review of f2fs filesystem * Limitation: Can only deal with power of 2 erase blocks * Limitation: Needs 6 concurrent erase blocks - low end SD cards might not have that many * Good review in LWN; yet to read complete source code 2. Linaro Connect planning * Need to shuffle around sessions due to limitation and travel logistics * 3 sessions proposed. ** f2fs review *** Need to confirm participation from Samsung, and whether any BPs will come out of the review. *** Good to have benchmarking done by Linaro folks (svenkatr -todo) ** eMMC4.6 features *** Needs Jedec membership, should be made as private / invite only meeting ** Swap on flash *** Will be a private session due to time constraints * Further discussions over email on Connect session scheduling / logistics. 3. Need to change the weekly meeting time to account for Friday holiday for Alex - to be discussed at Connect. Regards, Venkat.

12 years, 7 months

1
0
0 0

[ACTIVITY] (Rajanikanth H V) 2012-10-08 to 2012-10-12

by Rajanikanth HV

==== Activity Summary ==== * Completed implementing review comments from lee Jones, patch underwent enough optimization * Discussion and internal review with lee Jones regarding review comments * Spent some time on testing, observed expected interrupts from fuel-gauge driver ==== Plan ==== * Implementing Legacy platform data support to co-exist with DT bindings. ==== Issues ==== -- NA --

12 years, 7 months

1
0
0 0

[PATCH 0/3] A few cleanups and refactorings, sync w/ upstream

by Anton Vorontsov

Hello Pekka, Just a few updates to vmevents: - Some cleanups and refactorings -- needed for easier integration of 'memory pressure' work; - Forward to newer Linus' tree, fix conflicts. For convenience, the merge commit and all the patches can be pulled from this repo: git://git.infradead.org/users/cbou/linux-vmevent.git tags/vmevent-updates Thanks, Anton.

12 years, 7 months

2
6
0 0

[PATCH] mm: vmevent: Report number of used swap pages, not total amount of swap

by Anton Vorontsov

I think the actual intent was to report the number of used swap pages, not just total swap size. This patch fixes the issue. Signed-off-by: Anton Vorontsov <anton.vorontsov(a)linaro.org> --- mm/vmevent.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/mm/vmevent.c b/mm/vmevent.c index 1c2e72e..a059bed 100644 --- a/mm/vmevent.c +++ b/mm/vmevent.c @@ -46,7 +46,7 @@ static u64 vmevent_attr_swap_pages(struct vmevent_watch *watch, si_swapinfo(&si); - return si.totalswap; + return si.totalswap - si.freeswap; #else return 0; #endif -- 1.7.12.1

12 years, 7 months

1
0
0 0

[RFC 0/3] mm: vmevent: Stats accuracy improvements

by Anton Vorontsov

Hi all, Some time ago KOSAKI Motohiro noticed[1] that vmevent might be very inaccurate (up to 2GB inaccuracy on a very large machines) since per CPU stats synchronization happens either on time basis or when we hit stat thresholds. KOSAKI also told that perf API might be a good inspirations for further improvements, but I must admit I didn't fully get the idea, although I'm open to investigate this route too, but I guess it needs a bit more explanations. Also note that this is just an RFC, I just show some ideas and wonder how you feel about it. Since we now use memory pressure factor bolted into the reclaimer code path, we don't desperately need the accurate stats, but it's still nice thing to have/fix. Anyway, here we take two approaches: - Asynchronously sum vm_stat diffs and global stats. This is very similar to what we already have for per-zone stats, implemented in zone_page_state_snapshot(). The values still could be inaccurate, but overall this makes things better; - Implement configurable per CPU vmstat thresholds. This is much more powerful tool to get accurate statistics, but it comes with a price: it might cause some performance penalty as we'd update global stats more frequently (in a fast path), so users have to be careful. The two items are independent, so we might implement one or another, or both, or none, if desired. ;-) Thanks, Anton. p.s. Note that the patches are against my vmevent tree, i.e.: git://git.infradead.org/users/cbou/linux-vmevent.git [1] http://lkml.indiana.edu/hypermail/linux/kernel/1205.1/00062.html

12 years, 7 months

1
3
0 0

[ACTIVITY] (Linus Walleij) 2012-10-06 - 2012-10-12

by Linus Walleij

== Linus Walleij linusw == === Highlights === * Sent a second pull request to Torvalds for the v3.7 pinctrl patches that involved the simplified irqdomain. Torvalds pulled it in. * SPARSE_IRQ patch set ready to be pulled into ARM SoC when the merge window ends. Might need some rebasing and editing. * Back ported regmap MMIO from v3.5 to the internal v3.4 branch to help development of upstream kernel compliant features. * Resolved the issues around runtime PM / suspend and resume proper in the PL022 driver (and others). Planned a runtime PM session @connect with Ulf and Amit. * Reviewed the second iteration of Synaptics' RMI4 patch set. * Reviewed, slammed, got a v2 and eventually merged the pinctrl refactoring patches from Tomasz for the Samsung platforms. * Pulled in Jean-Christophe's AT91 pinctrl patch set, had it explode in next and request it to be rebased for inclusion past v3.7-rc1 or so. * Submitted and fixed up a patch deleting the "Nomadik" FSMC driver and replacing it with the one from SPEAr just named fsmc_nand.c. Fixed up breakage after a clash with a patch from Jean-Christophe. * Reviewed Vipin Kumar's FSMC patch set digging around details of userspace buffer mapping and why it doesn't work. * Reviewed a bunch of U9540 patches from Lee Jones internally before sending them out to the public lists. Debate continues on the lists. * Hunted some regressions and fixups, git them out the door. === Plans === * Test the PL08x patches on the Ericsson Research PB11MPCore and submit platform data for using pl08x DMA on that platform. * Look into other Ux500 stuff in need of mainlining... using an internal tracking sheet for this. * Look into regmap. Try something out, get to know it. === Issues === * Some internal time stealing, for example ST-Ericsson career development has taken some time this week. Thanks, Linus Walleij

12 years, 7 months

1
0
0 0

A15 and A7 big.Little support in different ways

by chao xie

hi For A15 and A7, there three kinds of usage model. 1. the cluster switch i found that there is the reference code about the switcher. So does linaro has any development based on it? 2. cpu migration It seems that only members can get the early code. Is that right? 3. MP mode I find a git "http://git.linaro.org/git/arm/big.LITTLE/mp.git". So does it contains the latest development of big and little system MP mode? Thanks.

12 years, 8 months

3
2
0 0

[ACTIVITY] (Anton Vorontsov) 2012-09-24 - 2012-10-05

by Anton Vorontsov

== Highlights == * Prepared and sent out v9 of KGDB NMI core patches, nowadays they all are in the mainline kernel; * All previously submitted persistent storage reworks and fixes are also in mainline now; * Sent out and discussed vmevents pressure factor. So far only Mel and John commented; * Prepared and sent refactorings for vmevents, another step to make it closer to the upstream. Also merged with the latest Linus' tree and requested Pekka to pull; == Plans == * Address Mel Gorman's implementation notes on vmevent's pressure factor. * Send out vm_stat enhancements for vmevent (need to prepare a good explanation); * Perform vmevents pressure factor use-case tests on a desktop, tune statistics (try Mel's idea of using multiple window sizes); * I wonder if it makes sense to resend the KGDB FIQ ARM and kiosk patches now, or maybe I should wait for the end of the merge window. I guess resending them now would be OK, we're halfway to the -rc1 anyways;

12 years, 8 months

1
0
0 0

[RFC] vmevent: Implement pressure attribute

by Anton Vorontsov

Hi all, This is just an RFC so far. It's an attempt to implement Mel Gorman's idea of detecting and measuring memory pressure by calculating the ratio of scanned vs. reclaimed pages in a given time frame. The implemented approach can notify userland about two things: - Constantly rising number of scanned pages shows that Linux is busy w/ rehashing pages in general. The more we scan, the more it's obvious that we're out of unused pages, and we're draining caches. By itself it's not critical, but for apps that want to maintain caches level (like Android) it's quite useful. The notifications are ratelimited by a specified amount of scanned pages. - Next, we calculate pressure using '100 - reclaimed/scanned * 100' formula. The value shows (in percents) how efficiently the kernel reclaims pages. If we take number of scanned pages and think of them as a time scale, then these percents basically would show us how much of the time Linux is spending to find reclaimable pages. 0% means that every page is a candidate for reclaim, 100% means that MM is not recliaming at all, it spends all the time scanning and desperately trying to find something to reclaim. The more time we're at the high percentage level, the more chances that we'll OOM soon. So, if we fail to find a page in a reasonable time frame, we're obviously in trouble, no matter how much reclaimable memory we actually have -- we're too slow, and so we'd better free something. Although it must be noted that the pressure factor might be affected by reclaimable vs. non-reclaimable pages "fragmentation" in an LRU. If there's a "hole" of reclaimable memory in an almost-OOMed system, the factor will drop temporary. On the other hand, it just shows how efficiently Linux is keeping the lists, it might be pretty inefficient, and the factor will show it. Some more notes: - Although the scheme sounds good, I noticed that reclaimer 'priority' level (i.e. scanning depth) better responds to pressure (it's more smooth), and so far I'm not sure how to make the original idea to work on a par w/ sc->priority level. - I have an idea, which I might want to try some day. Currently, the pressure callback is hooked into the inactive list reclaim path, it's the last step in the 'to be reclaimed' page's life time. But we could measure 'active -> inactive' migration speed, i.e. pages deactivation rate. Or we could measure inactive/active LRU size ratio, ideally behaving system would try to keep the ratio near 1, and it'll be close to 0 when inactive list is getting short (for anon LRU it'd be not 1, but zone->inactive_ratio actually). Thanks, Anton. --- include/linux/vmevent.h | 36 ++++++++++ mm/vmevent.c | 179 +++++++++++++++++++++++++++++++++++++++++++++++- mm/vmscan.c | 4 ++ 3 files changed, 218 insertions(+), 1 deletion(-) diff --git a/include/linux/vmevent.h b/include/linux/vmevent.h index b1c4016..1397ade 100644 --- a/include/linux/vmevent.h +++ b/include/linux/vmevent.h @@ -10,6 +10,7 @@ enum { VMEVENT_ATTR_NR_AVAIL_PAGES = 1UL, VMEVENT_ATTR_NR_FREE_PAGES = 2UL, VMEVENT_ATTR_NR_SWAP_PAGES = 3UL, + VMEVENT_ATTR_PRESSURE = 4UL, VMEVENT_ATTR_MAX /* non-ABI */ }; @@ -46,6 +47,11 @@ struct vmevent_attr { __u64 value; /* + * Some attributes accept two configuration values. + */ + __u64 value2; + + /* * Type of profiled attribute from VMEVENT_ATTR_XXX */ __u32 type; @@ -97,4 +103,34 @@ struct vmevent_event { struct vmevent_attr attrs[]; }; +#ifdef __KERNEL__ + +struct mem_cgroup; + +extern void __vmevent_pressure(struct mem_cgroup *memcg, + ulong scanned, + ulong reclaimed); + +static inline void vmevent_pressure(struct mem_cgroup *memcg, + ulong scanned, + ulong reclaimed) +{ + if (!scanned) + return; + + if (IS_BUILTIN(CONFIG_MEMCG) && memcg) { + /* + * The vmevent API reports system pressure, for per-cgroup + * pressure, we'll chain cgroups notifications, this is to + * be implemented. + * + * memcg_vm_pressure(target_mem_cgroup, scanned, reclaimed); + */ + return; + } + __vmevent_pressure(memcg, scanned, reclaimed); +} + +#endif + #endif /* _LINUX_VMEVENT_H */ diff --git a/mm/vmevent.c b/mm/vmevent.c index d643615..12d0131 100644 --- a/mm/vmevent.c +++ b/mm/vmevent.c @@ -4,6 +4,7 @@ #include <linux/vmevent.h> #include <linux/syscalls.h> #include <linux/workqueue.h> +#include <linux/interrupt.h> #include <linux/file.h> #include <linux/list.h> #include <linux/poll.h> @@ -30,6 +31,25 @@ struct vmevent_watch { wait_queue_head_t waitq; }; +struct vmevent_pwatcher { + struct vmevent_watch *watch; + struct vmevent_attr *attr; + struct vmevent_attr *samp; + struct list_head node; + + uint scanned; + uint reclaimed; + uint window; +}; + +static LIST_HEAD(vmevent_pwatchers); +static DEFINE_SPINLOCK(vmevent_pwatchers_lock); + +static uint vmevent_scanned; +static uint vmevent_reclaimed; +static uint vmevent_minwin = UINT_MAX; /* Smallest window in the list. */ +static DEFINE_SPINLOCK(vmevent_pressure_lock); + typedef u64 (*vmevent_attr_sample_fn)(struct vmevent_watch *watch, struct vmevent_attr *attr); @@ -141,6 +161,10 @@ static bool vmevent_match(struct vmevent_watch *watch) struct vmevent_attr *samp = &watch->sample_attrs[i]; u64 val; + /* Pressure is event-driven, not polled */ + if (attr->type == VMEVENT_ATTR_PRESSURE) + continue; + val = vmevent_sample_attr(watch, attr); if (!ret && vmevent_match_attr(attr, val)) ret = 1; @@ -204,6 +228,94 @@ static void vmevent_start_timer(struct vmevent_watch *watch) vmevent_schedule_watch(watch); } +static ulong vmevent_calc_pressure(struct vmevent_pwatcher *pw) +{ + uint win = pw->window; + uint s = pw->scanned; + uint r = pw->reclaimed; + ulong p; + + /* + * We calculate the ratio (in percents) of how many pages were + * scanned vs. reclaimed in a given time frame (window). Note that + * time is in VM reclaimer's "ticks", i.e. number of pages + * scanned. This makes it possible set desired reaction time and + * serves as a ratelimit. + */ + p = win - (r * win / s); + p = p * 100 / win; + + pr_debug("%s: %3lu (s: %6u r: %6u)\n", __func__, p, s, r); + + return p; +} + +static void vmevent_match_pressure(struct vmevent_pwatcher *pw) +{ + struct vmevent_watch *watch = pw->watch; + struct vmevent_attr *attr = pw->attr; + ulong val; + + val = vmevent_calc_pressure(pw); + + /* Next round. */ + pw->scanned = 0; + pw->reclaimed = 0; + + if (!vmevent_match_attr(attr, val)) + return; + + pw->samp->value = val; + + atomic_set(&watch->pending, 1); + wake_up(&watch->waitq); +} + +static void vmevent_pressure_tlet_fn(ulong data) +{ + struct vmevent_pwatcher *pw; + uint s; + uint r; + + if (!vmevent_scanned) + return; + + spin_lock(&vmevent_pressure_lock); + s = vmevent_scanned; + r = vmevent_reclaimed; + vmevent_scanned = 0; + vmevent_reclaimed = 0; + spin_unlock(&vmevent_pressure_lock); + + rcu_read_lock(); + list_for_each_entry_rcu(pw, &vmevent_pwatchers, node) { + pw->scanned += s; + pw->reclaimed += r; + if (pw->scanned >= pw->window) + vmevent_match_pressure(pw); + } + rcu_read_unlock(); +} +static DECLARE_TASKLET(vmevent_pressure_tlet, vmevent_pressure_tlet_fn, 0); + +void __vmevent_pressure(struct mem_cgroup *memcg, + ulong scanned, + ulong reclaimed) +{ + if (vmevent_minwin == UINT_MAX) + return; + + spin_lock_bh(&vmevent_pressure_lock); + + vmevent_scanned += scanned; + vmevent_reclaimed += reclaimed; + + if (vmevent_scanned >= vmevent_minwin) + tasklet_schedule(&vmevent_pressure_tlet); + + spin_unlock_bh(&vmevent_pressure_lock); +} + static unsigned int vmevent_poll(struct file *file, poll_table *wait) { struct vmevent_watch *watch = file->private_data; @@ -259,12 +371,40 @@ out: return ret; } +static void vmevent_release_pwatcher(struct vmevent_watch *watch) +{ + struct vmevent_pwatcher *pw; + struct vmevent_pwatcher *tmp; + struct vmevent_pwatcher *del = NULL; + int last = 1; + + spin_lock(&vmevent_pwatchers_lock); + + list_for_each_entry_safe(pw, tmp, &vmevent_pwatchers, node) { + if (pw->watch != watch) { + vmevent_minwin = min(pw->window, vmevent_minwin); + last = 0; + continue; + } + WARN_ON(del); + list_del_rcu(&pw->node); + del = pw; + } + + if (last) + vmevent_minwin = UINT_MAX; + + spin_unlock(&vmevent_pwatchers_lock); + synchronize_rcu(); + kfree(del); +} + static int vmevent_release(struct inode *inode, struct file *file) { struct vmevent_watch *watch = file->private_data; cancel_delayed_work_sync(&watch->work); - + vmevent_release_pwatcher(watch); kfree(watch); return 0; @@ -289,6 +429,36 @@ static struct vmevent_watch *vmevent_watch_alloc(void) return watch; } +static int vmevent_setup_pwatcher(struct vmevent_watch *watch, + struct vmevent_attr *attr, + struct vmevent_attr *samp) +{ + struct vmevent_pwatcher *pw; + + if (attr->type != VMEVENT_ATTR_PRESSURE) + return 0; + + if (!attr->value2) + return -EINVAL; + + pw = kzalloc(sizeof(*pw), GFP_KERNEL); + if (!pw) + return -ENOMEM; + + pw->watch = watch; + pw->attr = attr; + pw->samp = samp; + pw->window = (attr->value2 + PAGE_SIZE - 1) / PAGE_SIZE; + + vmevent_minwin = min(pw->window, vmevent_minwin); + + spin_lock(&vmevent_pwatchers_lock); + list_add_rcu(&pw->node, &vmevent_pwatchers); + spin_unlock(&vmevent_pwatchers_lock); + + return 0; +} + static int vmevent_setup_watch(struct vmevent_watch *watch) { struct vmevent_config *config = &watch->config; @@ -302,6 +472,7 @@ static int vmevent_setup_watch(struct vmevent_watch *watch) struct vmevent_attr *attr = &config->attrs[i]; size_t size; void *new; + int ret; if (attr->type >= VMEVENT_ATTR_MAX) continue; @@ -322,6 +493,12 @@ static int vmevent_setup_watch(struct vmevent_watch *watch) watch->config_attrs[nr] = attr; + ret = vmevent_setup_pwatcher(watch, attr, &attrs[nr]); + if (ret) { + kfree(attrs); + return ret; + } + nr++; } diff --git a/mm/vmscan.c b/mm/vmscan.c index 99b434b..f4dd1e0 100644 --- a/mm/vmscan.c +++ b/mm/vmscan.c @@ -20,6 +20,7 @@ #include <linux/init.h> #include <linux/highmem.h> #include <linux/vmstat.h> +#include <linux/vmevent.h> #include <linux/file.h> #include <linux/writeback.h> #include <linux/blkdev.h> @@ -1334,6 +1335,9 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec, nr_scanned, nr_reclaimed, sc->priority, trace_shrink_flags(file)); + + vmevent_pressure(sc->target_mem_cgroup, nr_scanned, nr_reclaimed); + return nr_reclaimed; } -- 1.7.12.1

12 years, 8 months

3
5
0 0

Kernel Storage Team weekly meeting mintues (Oct 5, 2012)

by Venkatraman S

Attendees: Arnd Bergmann, Balaji TK, Jakub Pavelek, Luca Porzio, Venkatraman S (#linaro-storage: arnd, balajitk, Xruxa, lupo, svenkatr) Discussion notes:- 1. New members joining storage team - Balaji TK and Ulf Hansson 2. Linaro Connect - attendance:- All above are planning to attend except Venkatraman S 3. Proposed topics to discuss at Connect and track leads a) Kernel changes for eMMC4.6 support (Lupo ) b) ext4 optimizations for Flash (Arnd) c) Swap on Flash optimizations (Venkatraman S) d) review of Samsung's F2FS file system ( Samsung / Arnd) 4. Possibility to include SDIO v3 features in next cycle ? To check ! 5. Discussion about eMMC4.6 FOTA feature and possible ways to implement Is it really a Kernel feature (as eMMC will become offline while updating) ? Need to study SCSI / block subsystem's methods and take a similar approach (IOCTL). 6. "ext4 move journal to enhanced area" BP is completed. Performance comparison to be published and the BP to be closed. - (Balaji TK to close) 7. core eMMC4.5 features to be reviewed and upstreamed with community support. Need to focus on core FS / Block layer optimization for flash. But still some features need attention (context ID and Power Off Notify) Regards, Venkat.

12 years, 8 months

1
0
0 0

[ACTIVITY] (Deepak Saxena) Sept 30th - Oct 5

by Deepak Saxena

== Deepak Saxena <dsaxena> == === Highlights === * Started creating Connect sessions and fleshing out plans * Looking at aarch64 SOC code, submitted a patch (!), and provided some feedback * Continued work on cleaning up existing BPs * Working on breaking down single zImage work into BPs * More changes to https://wiki.linaro.org/WorkingGroups/Kernel/BlueprintProcess === Plans === * Finish Connection session creation * Work with Jakub and engineers to fill in details of single zImage BPs * Create some new roadmap cards: - Existing Android BPs that are actively being worked on w/o a card - Enabling of EMMC-4.6 Features - EMMC Power Management == Issues === === Travel/Time Off === * Linaro US Holiday Monday Oct 8th * Off Friday Oct, 19th * Connect Copenhagen

12 years, 8 months

1
0
0 0

[ACTIVITY] (John Stultz) Oct 1-5th

by John Stultz

=== Highlights === * Pinged tglx on maybe reducing my 3.7 queue and possibly pushing directly to -next in the future to improve testing * Briefly chatted with Zach on arm64 android * Had to deal with internal infrastructure outage some this week. * Responded to some limited comments on my volatile range patches that I sent out * Sent Taras Glek some pointers to other work being done in Linaro that is relevant to Mozilla * Reviewed some paravirt-clock patches, suggested alternative methods to implement them. * Assisted Matheiu on some questions wrt sending his patchset to akpm * Sent Android subteam email & moved next weeks meeting to google hangouts instead of mumble * Started looking at madvise() volatile ranges === Plans === * Hopefully paternity leave? === Issues === * NA

12 years, 8 months

1
0
0 0

[ACTIVITY] (Linus Walleij) 2012-09-29 - 2012-10-05

by Linus Walleij

== Linus Walleij linusw == === Highlights === * Sent pull request to Torvalds for the v3.7 GPIO patch queue. Torvalds pulled it in. * Sent pull request to Torvalds for the v3.7 pinctrl patch queue. Torvalds pulled it in. * Rob Herring ACK:ed my patch to irqdomains to extend the simple domain. Still seeking an ACK from Grant. * SPARSE_IRQ patch set ready to be pulled into ARM SoC when the merge window ends. * Created and edited a few blueprints trying to follow the new process. * Strong progress on pinctrl adaption internally at ST-Ericsson. Helped out with review and merge. === Plans === * Discussed handling for runtime PM and ordinary suspend/resume in the PL022 SPI driver, discussion ongoing with Ulf Hansson. I'm a bit confused. * Test the PL08x patches on the Ericsson Research PB11MPCore and submit platform data for using pl08x DMA on that platform. * Look into other Ux500 stuff in need of mainlining... using an internal tracking sheet for this. * Look into regmap. Try something out, get to know it. === Issues === * Some internal time stealing, for example ST-Ericsson career development has taken some time this week. Thanks, Linus Walleij

12 years, 8 months

1
0
0 0

[ACTIVITY] (John Stultz) Sept 24-28th

by John Stultz

Gotten in the bad habit of working until late on Friday and then just dropping offline and forgetting to send out status. === Highlights === * Sent my 3.7 queue to Thomas, unfortunately he's not responded. * Had dinner with Taras Glek of Mozilla to discuss the volatile range changes and further details of how Mozilla uses Android functionality both in Android as well as in FirefoxOS. * Helped Mark in summarizing my role with Android Upstreaming. * Ran the bi-weekly Upstreaming meeting on mumble * Walked Jakub through some of the Android Upstreaming blueprints as well as some of the history in Linaro's processes. * Scratched out some more strategy thoughts on Linaro's roll in Android upstreaming, and send the draft to Deepak. * Implemented SIGBUS on volatile range access. * Wrote up a large summary document on volatile ranges * Sent out both v7 volatile range patches (including SIGBUS functoinality) along with the long summary document to lkml === Plans === * Waiting for a baby * Look at details on madvise() based volatile ranges on anonymous memory, per Mozilla's preference * 3.7 merge window work === Issues === * Some internal infrastructure snafus catching me atm. Hopefully won't last too long.

12 years, 8 months

1
0
0 0

[ACTIVITY] (Linus Walleij) 2012-09-24 - 2012-09-28

by Linus Walleij

== Linus Walleij linusw == === Highlights === * Finally had the RCU lockup issue fixed - Paul McKenney found the problem and fixed it. It turns out the same bug has been seen on the OMAP, where Paul Walmsley, Becy Bruce et al had reported it. http://marc.info/?l=linux-kernel&m=134749030206016&w=2 Hunting and testing this took a lot of time. * To progress on SPARSE_IRQ I proposed a patch to remove the warning from the GIC, and alerted Rob Herring. He didn't like that but proposed to fix the problem by fixing all descriptor allocation, which will take more time. * Proposed a patch fixing the pinctrl IRQ desc allocations (these happen really early) and was adviced to introduce a linear IRQ domain, did this, was proposed to refactor IRQ domain core, proposed a patch for this, now following up. * Found out how to really get rid of the ugly workarounds in the PL011 driver, only a small ST-Micro derivate-only patch was needed in the end to adjust baud rate down with ~5% on high baud rates. The rest of the patches were bogus, we weren't interfacing the TTY layer correctly. Alan Cox told me how to do it properly and lo and behold, it just works now. (All work synchronized back to ST-Ericsson's internal codebase.) * Sent out a pinctrl patch for Nomadik I2C, but still pondering this due to regressions. * Discussed resource handling for runtime PM and ordinary suspend/resume in the PL022 SPI driver, discussion ongoing. * Reviewed and merged various GPIO patches. * Reviewed and merged various pinctrl patches. Pondering PM semantics. * Reviewed some ux500 patches coming from Lee. === Plans === * Ongoing work on sparse IRQ for Nomadik and Ux500 with the required IRQ domain refactorings. * Follow up the rest of the above. * Test the PL08x patches on the Ericsson Research PB11MPCore and submit platform data for using pl08x DMA on that platform. * Look into other Ux500 stuff in need of mainlining... using an internal tracking sheet for this. * Look into regmap. Try something out, get to know it. === Issues === * N/A Thanks, Linus Walleij

12 years, 8 months

1
0
0 0

[PATCH v9 0/11] KGDB/KDB FIQ (NMI) debugger

by Anton Vorontsov

Hi all, Here comes v9... Greg, patches 1-7 have gotten all the needed acks from Alan and Jason, so it would be awesome if you could apply them to your tty-next tree. ARM patches still need Russell's acks, but these can wait, plus if the core support hits v3.7, it would be much easier to handle the rest (including the patches that add "kiosk mode" support for KDB). Thanks! In v9: - KGDB patches got their Acked-by tags from Jason Wessel; - A small 64 bit fixup for kgdb_nmi driver, thanks to Jason; - Yet again rebased on the latest and greatest tty-next, just to make sure everything applies cleanly. These patches can be found in the following repo (based on tty-next): git://git.infradead.org/users/cbou/linux-nmi-kdb.git master Old changelogs and rationale for these patches can be found here: v1-v5, rationale: http://lkml.org/lkml/2012/9/10/2 v6: http://lkml.org/lkml/2012/9/10/2 v7: http://lkml.org/lkml/2012/9/13/367 v8: http://lkml.org/lkml/2012/9/19/525 -- arch/arm/Kconfig | 19 ++ arch/arm/common/vic.c | 28 ++ arch/arm/include/asm/hardware/vic.h | 2 + arch/arm/include/asm/kgdb.h | 8 + arch/arm/kernel/Makefile | 1 + arch/arm/kernel/entry-armv.S | 167 +----------- arch/arm/kernel/entry-header.S | 170 ++++++++++++ arch/arm/kernel/kgdb_fiq.c | 124 +++++++++ arch/arm/kernel/kgdb_fiq_entry.S | 87 +++++++ arch/arm/mach-versatile/Makefile | 1 + arch/arm/mach-versatile/kgdb_fiq.c | 31 +++ drivers/tty/serial/Kconfig | 19 ++ drivers/tty/serial/Makefile | 1 + drivers/tty/serial/amba-pl011.c | 73 +++++- drivers/tty/serial/kgdb_nmi.c | 402 +++++++++++++++++++++++++++++ drivers/tty/serial/kgdboc.c | 9 + drivers/tty/serial/serial_core.c | 17 ++ include/linux/kdb.h | 29 ++- include/linux/kgdb.h | 13 + include/linux/serial_core.h | 1 + kernel/debug/debug_core.c | 14 +- kernel/debug/kdb/kdb_main.c | 31 +++ 22 files changed, 1054 insertions(+), 193 deletions(-)

12 years, 8 months

3
15
0 0

[ACTIVITY] (Anton Vorontsov) 2012-09-17 - 2012-09-21

by Anton Vorontsov

== Highlights == * Performed some proof-of-concept tests on the Mel Gorman's idea of using nr_scanned/(nr_reclaimed+1) ratio for memory pressure detection -- and the idea seems to work great. Didn't have time to implement it in the vmevent API, but hopefully this week I'll have something to show; * Addressed Jason Wessel's comments on KDB FIQ debugger patches, i.e. now we only register ttyNMI interface and disable_nmi command on arches that actually support NMI rerouting, plus reworked enable_nmi callback; * Got a bunch of new ideas from Android folks on KDB FIQ debugger, cpufreq, low memory handling, and more. Lots of things to investigate and implement; * Some power supply and pstore maintenance activity, reviewed over 90 patches in total.

12 years, 8 months

1
0
0 0

[PATCH v8 0/11] KGDB/KDB FIQ (NMI) debugger

by Anton Vorontsov

Hi all, In v8, addressed Jason's comments: - Changed kgdb_enable_nmi() weak function to kgdb_arch callbck; - We no longer register disable_nmi command if arch does not register KGDB NMI handling (i.e. not filling kgdb_arch.enable_nmi callback); - The same is for ttyNMI: if architecure does not provide us with enable_nmi call, we don't need the tty device. Of course, there is no way to tell wether a specific serial device can be used for NMI debugging, as it is not serial-device specific, but specific to whether IRQ can be rerouted to an NMI (for most our cases, pretty much every IRQ can be rerouted, e.g. a hot-pluggable serial device on a PCI bus). - Rebased on the latest and greatest tty-next, just in case. These patches can be found in the following repo (based on tty-next): git://git.infradead.org/users/cbou/linux-nmi-kdb.git master Old changelogs and rationale for these patches can be found here: v1-v5, rationale: http://lkml.org/lkml/2012/9/10/2 v6: http://lkml.org/lkml/2012/9/10/2 v7: http://lkml.org/lkml/2012/9/13/367 Thanks, -- arch/arm/Kconfig | 19 ++ arch/arm/common/vic.c | 28 +++ arch/arm/include/asm/hardware/vic.h | 2 + arch/arm/include/asm/kgdb.h | 8 + arch/arm/kernel/Makefile | 1 + arch/arm/kernel/entry-armv.S | 167 +------------ arch/arm/kernel/entry-header.S | 170 +++++++++++++ arch/arm/kernel/kgdb_fiq.c | 124 +++++++++ arch/arm/kernel/kgdb_fiq_entry.S | 87 +++++++ arch/arm/mach-versatile/Makefile | 1 + arch/arm/mach-versatile/kgdb_fiq.c | 31 +++ drivers/tty/serial/Kconfig | 19 ++ drivers/tty/serial/Makefile | 1 + drivers/tty/serial/amba-pl011.c | 73 +++++- drivers/tty/serial/kgdb_nmi.c | 402 ++++++++++++++++++++++++++++++ drivers/tty/serial/kgdboc.c | 9 + drivers/tty/serial/serial_core.c | 17 ++ include/linux/kdb.h | 29 ++- include/linux/kgdb.h | 13 + include/linux/serial_core.h | 1 + kernel/debug/debug_core.c | 14 +- kernel/debug/kdb/kdb_main.c | 31 +++ 22 files changed, 1054 insertions(+), 193 deletions(-)

12 years, 8 months

2
16
0 0

[PATCH] arm: add SDIV/UDIV emulation for ARMv7 processors

by Vladimir Murzin

The Virtualization Extensions introduce the requirement for an ARMv7-A implementation to include SDIV and UDIV. Any implementation of the Virtualization Extensions must include the SDIV and UDIV instructions in the Thumb and ARM instruction sets. In an ARMv7-A implementation that does not include the Virtualization Extensions, it is IMPLEMENTATION DEFINED whether: * SDIV and UDIV are not implemented * SDIV and UDIV are implemented only in the Thumb instruction set * SDIV and UDIV are implemented in the Thumb and ARM instruction sets. This patch adds a handler to trap and emulate unimplemented SDIV and UDIV instructions in ARM and Thumb modes. Also some basic statistic is exported via /proc/cpu/idiv_emulation Signed-off-by: Vladimir Murzin <murzin.v(a)gmail.com> --- arch/arm/kernel/Makefile | 1 + arch/arm/kernel/idiv_emulate.c | 194 ++++++++++++++++++++++++++++++++++++++++ arch/arm/mm/Kconfig | 26 ++++++ 3 files changed, 221 insertions(+), 0 deletions(-) create mode 100644 arch/arm/kernel/idiv_emulate.c diff --git a/arch/arm/kernel/Makefile b/arch/arm/kernel/Makefile index 7ad2d5c..057a630 100644 --- a/arch/arm/kernel/Makefile +++ b/arch/arm/kernel/Makefile @@ -62,6 +62,7 @@ obj-$(CONFIG_OF) += devtree.o obj-$(CONFIG_CRASH_DUMP) += crash_dump.o obj-$(CONFIG_SWP_EMULATE) += swp_emulate.o CFLAGS_swp_emulate.o := -Wa,-march=armv7-a +obj-$(CONFIG_IDIV_EMULATE) += idiv_emulate.o obj-$(CONFIG_HAVE_HW_BREAKPOINT) += hw_breakpoint.o obj-$(CONFIG_CPU_XSCALE) += xscale-cp0.o diff --git a/arch/arm/kernel/idiv_emulate.c b/arch/arm/kernel/idiv_emulate.c new file mode 100644 index 0000000..f021d0e --- /dev/null +++ b/arch/arm/kernel/idiv_emulate.c @@ -0,0 +1,195 @@ +/* + * linux/arch/arm/kernel/idiv_emulate.c + * + * This code is based on swp_emulate.c + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + * + * Implements emulation of the SDIV/UDIV instructions. They are defined for + * ARMv7-R and ARMv7-M profiles (in Thumb state only) and are UNDEFINED in the + * ARMv7-A profile. SDIV/UDIV are present by default on Cortex-A15. + * + * This emulation allow using integer divide instructions in case hardware is + * not presented + * + * Syntax of SDIV/UDIV instructions: SDIV/UDIV<c> <Rd>,<Rn>,<Rm> + * Where: Rd = the destination register. + * Rn = the register that contains the dividend. + * Rm = the register that contains the divisor. + */ + +#include <linux/init.h> +#include <linux/kernel.h> +#include <linux/proc_fs.h> +#include <linux/sched.h> +#include <linux/syscalls.h> +#include <linux/perf_event.h> + +#include <asm/opcodes.h> +#include <asm/traps.h> +#include <asm/uaccess.h> + +/* + * Macros/defines for extracting register numbers from instruction. + */ +#define EXTRACT_REG_NUM(instruction, offset) \ + (((instruction) & (0xf << (offset))) >> (offset)) + +/* + * Offsets for ARM mode + */ +#define RD_OFFSET 16 +#define RN_OFFSET 0 +#define RM_OFFSET 8 + +/* + * Offsets for Thumb mode + */ +#define RD_T_OFFSET 8 +#define RN_T_OFFSET 16 +#define RM_T_OFFSET 0 + +#define TYPE_OF_DIV (1 << 21) + +static unsigned long sdivcounter; +static unsigned long udivcounter; +static pid_t previous_pid; + +#ifdef CONFIG_PROC_FS +static int proc_read_status(char *page, char **start, off_t off, int count, + int *eof, void *data) +{ + char *p = page; + int len; + + p += sprintf(p, "Emulated UDIV:\t\t%lu\n", udivcounter); + p += sprintf(p, "Emulated SDIV:\t\t%lu\n", sdivcounter); + if (previous_pid != 0) + p += sprintf(p, "Last process:\t\t%d\n", previous_pid); + + len = (p - page) - off; + if (len < 0) + len = 0; + + *eof = (len <= count) ? 1 : 0; + *start = page + off; + + return len; +} +#endif + +static u32 emulate_udiv(u32 n, u32 base) +{ + udivcounter++; + + return n/base; +} + +static s32 emulate_sdiv(s32 n, s32 base) +{ + sdivcounter++; + + return n/base; +} + +static int idiv_handler(struct pt_regs *regs, unsigned int instr) +{ + long dividend, divisor, dest, res; + + perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, regs, regs->ARM_pc); + + res = arm_check_condition(instr, regs->ARM_cpsr); + + switch (res) { + case ARM_OPCODE_CONDTEST_PASS: + break; + case ARM_OPCODE_CONDTEST_FAIL: + /* Condition failed - return to next instruction */ + regs->ARM_pc += 4; + return 0; + case ARM_OPCODE_CONDTEST_UNCOND: + if (!thumb_mode(regs)) + return -EFAULT; + break; + default: + return -EINVAL; + } + + if (current->pid != previous_pid) { + pr_debug("\"%s\" (%ld) uses idiv instruction\n", + current->comm, (unsigned long)current->pid); + previous_pid = current->pid; + } + + if (!thumb_mode(regs)) { + dividend = regs->uregs[EXTRACT_REG_NUM(instr, RN_OFFSET)]; + divisor = regs->uregs[EXTRACT_REG_NUM(instr, RM_OFFSET)]; + dest = EXTRACT_REG_NUM(instr, RD_OFFSET); + } else { + dividend = regs->uregs[EXTRACT_REG_NUM(instr, RN_T_OFFSET)]; + divisor = regs->uregs[EXTRACT_REG_NUM(instr, RM_T_OFFSET)]; + dest = EXTRACT_REG_NUM(instr, RD_T_OFFSET); + } + +/* + * In an ARMv7-A profile implementation that supports the SDIV and UDIV + * instructions, divide-by-zero always returns a zero result. + * In fact, integer division emulation provided by gcc lib has already handle + * division by zero case sending the signal to the caused process. Emulate this + * behavior here as well. + */ + if (!divisor) { + siginfo_t info; + + info.si_code = FPE_INTDIV; + info.si_signo = SIGFPE; + info.si_errno = 0; + + arm_notify_die("Division by zero", regs, &info, 0, 0); + + goto out; + } + + if (instr & TYPE_OF_DIV) + res = emulate_udiv((u32)dividend, (u32)divisor); + else + res = emulate_sdiv((s32)dividend, (s32)divisor); + + regs->ARM_pc += 4; + regs->uregs[dest] = res; + +out: + return 0; +} + +static struct undef_hook idiv_hook = { + .instr_mask = 0x0310f010, + .instr_val = 0x0310f010, + .cpsr_mask = MODE_MASK, + .cpsr_val = USR_MODE, + .fn = idiv_handler +}; + +static int __init idiv_emulation_init(void) +{ +#ifdef CONFIG_PROC_FS + struct proc_dir_entry *res; + + res = create_proc_entry("cpu/idiv_emulation", S_IRUGO, NULL); + + if (!res) + return -ENOMEM; + + res->read_proc = proc_read_status; +#endif /* CONFIG_PROC_FS */ + + pr_notice("Registering SDIV/UDIV emulation handler\n"); + + register_undef_hook(&idiv_hook); + + return 0; +} + +late_initcall(idiv_emulation_init); diff --git a/arch/arm/mm/Kconfig b/arch/arm/mm/Kconfig index 101b968..8e2fa94 100644 --- a/arch/arm/mm/Kconfig +++ b/arch/arm/mm/Kconfig @@ -651,6 +651,32 @@ config SWP_EMULATE If unsure, say Y. +config IDIV_EMULATE + bool "Emulate SDIV/UDIV instructions" + depends on !CPU_USE_DOMAINS && CPU_V7 + select HAVE_PROC_CPU if PROC_FS + default y if SMP + help + + The Virtualization Extensions introduce the requirement for an + ARMv7-A implementation to include SDIV and UDIV. Any + implementation of the Virtualization Extensions must include + the SDIV and UDIV instructions in the Thumb and ARM + instruction sets. + + In an ARMv7-A implementation that does not include the + Virtualization Extensions, it is IMPLEMENTATION DEFINED whether: + * SDIV and UDIV are not implemented + * SDIV and UDIV are implemented only in the Thumb instruction set + * SDIV and UDIV are implemented in the Thumb and ARM + instruction sets. + + This option allows to handle exeptions due to unimplemented + SDIV and UDIV instructions in ARM and Thumb modes. + + If unsure, say Y. + + config CPU_BIG_ENDIAN bool "Build big-endian kernel" depends on ARCH_SUPPORTS_BIG_ENDIAN -- 1.7.8.6

12 years, 8 months

3
2
0 0

[PATCH v7 0/11] KGDB/KDB FIQ (NMI) debugger

by Anton Vorontsov

Hi all, Here comes the lucky v7: - Per Alan Cox's suggestion added hangup method and removed a small leftover; - Per Colin Cross' suggestion moved IRQ quiescing logic into poll_get_char routine. IIUC, Alan is less unhappy about it. As a result, clear_irq() callback dropped. These patches can be found in the following repo (based on tty-next): git://git.infradead.org/users/cbou/linux-nmi-kdb.git master Old changelogs and rationale for these patches can be found here: v1-v5, rationale: http://lkml.org/lkml/2012/9/10/2 v6: http://lkml.org/lkml/2012/9/10/2 Thanks, -- arch/arm/Kconfig | 19 ++ arch/arm/common/vic.c | 28 ++ arch/arm/include/asm/hardware/vic.h | 2 + arch/arm/include/asm/kgdb.h | 8 + arch/arm/kernel/Makefile | 1 + arch/arm/kernel/entry-armv.S | 167 +----------- arch/arm/kernel/entry-header.S | 170 +++++++++++++ arch/arm/kernel/kgdb_fiq.c | 99 ++++++++ arch/arm/kernel/kgdb_fiq_entry.S | 87 +++++++ arch/arm/mach-versatile/Makefile | 1 + arch/arm/mach-versatile/kgdb_fiq.c | 31 +++ drivers/tty/serial/Kconfig | 19 ++ drivers/tty/serial/Makefile | 1 + drivers/tty/serial/amba-pl011.c | 73 +++++- drivers/tty/serial/kgdb_nmi.c | 396 +++++++++++++++++++++++++++++ drivers/tty/serial/kgdboc.c | 6 + drivers/tty/serial/serial_core.c | 17 ++ include/linux/kdb.h | 29 ++- include/linux/kgdb.h | 33 +++ include/linux/serial_core.h | 1 + kernel/debug/debug_core.c | 36 ++- kernel/debug/kdb/kdb_main.c | 29 +++ 22 files changed, 1060 insertions(+), 193 deletions(-)

12 years, 8 months

5
26
0 0

[ACTIVITY] (John Stultz) Sept 10-14

by John Stultz

Whoops. Sorry this is late. === Highlights === * Attended a connect planning meeting * Sent out first pass of refinied jiffies patch (help kill CLOCK_TICK_RATE for unified-zImage) * Sent out half of the -stable backports * Debugged and fixed a 32bit time regression in 3.6-rc, got it merged in tip for upstream. * Arve pinged me on an issue they were seeing with the upstream alarmtimer code. Implemented a solution and will be sending it out for discussion shortly. * Discussed with Mel Gorman on his approach to low-memory notification * Emailed Colin Cross and AntonV to get a coffee meeting scheduled. === Plans === * Make first pass at sending SIGBUS on accessing purged volatile pages * Send out alarmtimer fix for discussion * Send out vsyscall update patches for discussion * Finish testing and sending -stable backports === Issues === * NA

12 years, 8 months

1
0
0 0

(no subject)

by Deepak Saxena

== Deepak Saxena <dsaxena> == === Highlights === * Got mostly caught up on all email, minus upstream mailing lists * Had initial discussions with Jakub on project management * OPSCOM roadmap card review === Plans === * Update roadmap cards as per OPSCOM review * Go through all old blueprints and clean them up * Create BPs for currently untracked work (single zImage, aarch64) * Catch up on Linux Plumber's Conf and Kernel Summit presentations * Write up blueprint process for project status tracking * Get caught up on linux-arm-kernel list === Issues === === Travel/Time Off === * Travelling to NYC for 1 week,Sept 22nd - 28th, will be working half time * Copenhagen Connect

12 years, 8 months

1
0
0 0

[Activity] (Omar Ramirez) Sep 10 - Sep 16 2012

by Omar Ramirez Luna

=== omarrmz === === Highlights === * Reworked and resubmitted iommu series: DONE. * Submitted mailbox DT series: DONE. * AARCH64: - Setup git.linaro.org credentials: DONE - Hosted a mirror branch of aarch64-linux with versatile code: DONE === Plans === * Track iommu and mailbox patches into upstream. * Submit bootctrl patch to tidspbridge. * Continue Aarch64 work. Regards, Omar

12 years, 8 months

1
0
0 0

[ACTIVITY] (Thomas Abraham) Sept-10

by Thomas Abraham

== Status == * Started working on common clock support for Exynos4 platforms. == Next Week == * Complete initial implementation of the common clock support for Exynos4 platforms. == Misc == * Was on leave from Sept-11 to Sept-14.

12 years, 8 months

1
0
0 0

[ACTIVITY] (Dong Aisheng) Sept-10-Sept-14

by Dong Aisheng

=== Highlights === * Setup kvm development environment. 1) build Cortex A15x1 fast model simulator 2) Successfully ran A15x1 simulator on ubuntu PC 3) Ran A15 kernel via qemu-system-arm within simulator environment * Research kvm subsystem and qemu (read doc and code) * Research big.LITTLE tech, read ARM doc on it and some online technical articles on switcher and bit.LITTLE MP scheduler, started to read ARM reference switcher code from: http://git.linaro.org/gitweb?p=arm/big.LITTLE/switcher.git;a=summary

12 years, 8 months

1
0
0 0

[ACTIVITY] (Anton Vorontsov) 2012-09-03 - 2012-09-14

by Anton Vorontsov

== Highlights == * Lots of work for KDB FIQ/NMI debugger, sent out v5, v6 and v7. Switched NMI console driver to tty_port infrastructure, discussed the ways to handle serial interrupts quiescing, made magic sequence configurable. TTY patches got Alan Cox's ack. * Looked into timer slack support for cgroups. While it's indeed somewhat close to deferrable timers, it's still not exactly what we need. Timer slack let us group wakeups, but there is still no way to tell that the timer/poll/select itself should not cause wakeups. As for deferrable timers, there are still no comments. * Looked into ideas on early OOM detection from the userland. There are a lot of inspirational thoughts on measuring memory stress levels, and I definitely need to perform some tests. I'll prepare a detailed answer to the ideas, but the most interesting part is that the ideas are based on vmstat monitoring, which is important: if we're going to use vmstat, we'll eventually get beaten by vmstat downsides: slow/expensive updates (so the latency if the pre-OOM detection would be equal to vmstat_update_time). And for Android low memory killer use case, pre-OOM condition is important, but not the main problem: the main "problem" is to monitor memory level when there is no stress at all (but deferrable timers would solve this). * Rebased and applied pstore/ftrace debugfs rework, this is now in linux-next. Also reviewed/fixed FDT support for pstore, by Bryan Freed @ Chromium.org.

12 years, 8 months

1
0
0 0

[PATCH/RFC 0/4] Deferrable timers support for timerfd API

by Anton Vorontsov

Hi all, This patch set implements a userland-side API for generic deferrable timers, per linux/timer.h: * A deferrable timer will work normally when the system is busy, but * will not cause a CPU to come out of idle just to service it; instead, * the timer will be serviced when the CPU eventually wakes up with a * subsequent non-deferrable timer. These timers are crucial for power saving, i.e. periodic tasks that want to work in background when the system is under use, but don't want to cause wakeups themselves. The deferred timers are somewhat orthogonal to high-res external timers, since the deferred timer is tied to the system load, not just to some external decrementer source. So, currently, the implementation has a HZ precision, and the maximum interval is jiffies resolution (i.e. with HZ=1000, on 32 bit that would be around max 49 days). Of course we can implement longer timeouts by rearming the timer, although it probably wouldn't make much sense in real world, so we keep it simple and just return E2BIG if we don't like the interval. Note that the code is still using time calculation that is done by the hrtimer routines, so we pretty much reuse everything except for the timer events themselves (i.e. we use calculation results of hrtimer_forward_now() and hrtimer_expires_remaining(), but never start the hrtimer). So the code path is pretty much the same for both hrtimers and deferrable timers. We will use the timers to periodically read /proc/vmstat without forcibly waking up the system; but let's see, maybe there are other use cases that might be interesting for PM folks. Thanks! Anton. -- fs/timerfd.c | 87 +++++++++++++++++++++++++++++++++++++++++++------ include/linux/jiffies.h | 3 ++ include/linux/ktime.h | 3 +- include/linux/timerfd.h | 4 ++- kernel/time.c | 23 +++++++++++++ 5 files changed, 108 insertions(+), 12 deletions(-)

12 years, 8 months

2
5
0 0

[ACTIVITY] (Linus Walleij) 2012-09-08 - 2012-09-14

by Linus Walleij

== Linus Walleij linusw == === Highlights === * Finally sent out a patch for SPARSE_IRQ on the Nomadik and ux500. Sadly it seems that to get a kernel that isn't screaming at you, transitioning to sparse IRQ can not be done unless you move over to irqdomains, all dynamic IRQ numbers and device tree at the same time. This puzzled me for a long time. * Started to drill into a RCU lockup regression that is still in the v3.6-rc series, but I'm uncertain if it only affects ux500 or if it's a generic problem. John Stultz has seen the same message. * Helped Benjamin with pre-review of a patch to genalloc. * First review of Lee's patch set for HREF/MOP500 device tree support for the parts that I understand. * Fixed a smallish regression on the ks8695 watchdog that came as fallout from the timer patches. * Reviewed and merged various GPIO patches. * Reviewed and merged various pinctrl patches. Currently the Marvell patch set is taking up a lot of bandwidth but starting to look really nice. * Attended an internal workshop on kernel architecture that stole away some three working days or so, but hey, we get to influence the ST-Ericsson internal kernel. === Plans === * Ongoing work on sparse IRQ for Nomadik and Ux500. * Test the PL08x patches on the Ericsson Research PB11MPCore and submit platform data for using pl08x DMA on that platform. * Look into other Ux500 stuff in need of mainlining... using an internal tracking sheet for this. * Look into regmap. Try something out, get to know it. === Issues === * N/A Thanks, Linus Walleij

12 years, 8 months

1
0
0 0

3.1.5 kernel on Pandaboard

by Tirtha Ghosh

Hi All, I was trying booting 3.1.5 linaro kernel with wheezy debian filesystem on pandaboard, but I found ethernet is not working. On further check I found, there is no net interface under /sys/class/net/ except loopback. Though I have checked kernel config is having scm91x (and scm911x) enabled. I am really not sure about net driver details but I could not see these drivers are getting probed at kernel boot. Whereas, I have taken 3.2 kernel from kernel.org and it worked perfectly fine and there is no change in defconfig file from 3.1.5. I also observed the code structure for ethernet drivers are different between 3.1.5 and 3.2 but driver files are same except *ndo_set_rx_mode* has been implemented in 3.2 in place of * ndo_set_multicast_list.* any idea, about this problem? is it known??? Or is it something I am missing?? Thanks in advance -- With Best Regards, *Tirtha* [image: Nico Nico animated GIF --]

12 years, 8 months

3
4
0 0

3.1.5 kernel on Pandaboard

by Tirtha Ghosh

Hi All, I was trying booting 3.1.5 linaro kernel with wheezy debian filesystem on pandaboard, but I found ethernet is not working. On further check I found, there is no net interface under /sys/class/net/ except loopback. Though I have checked kernel config is having scm91x (and scm911x) enabled. I am really not sure about net driver details but I could not see these drivers are getting probed at kernel boot. Whereas, I have taken 3.2 kernel from kernel.org and it worked perfectly fine and there is no change in defconfig file from 3.1.5. I also observed the code structure for ethernet drivers are different between 3.1.5 and 3.2 but driver files are same except *ndo_set_rx_mode* has been implemented in 3.2 in place of * ndo_set_multicast_list.* any idea, about this problem? is it known??? Or is it something I am missing?? Thanks in advance -- With Best Regards, *Tirtha* [image: Nico Nico animated GIF --]

12 years, 8 months

1
0
0 0

[PATCH v6 0/12] KGDB/KDB FIQ (NMI) debugger

by Anton Vorontsov

Hi all, Here is a new revision, mostly tty reworks. The new tty_port stuff is a bliss: no more per-driver mutex, no more counting for open(), well-separated initialization callbacks (I hope I got them right :-). But since I now use a lot of new tty_port stuff, I had to rebase the patch set on top of tty-next, so there's no point in cherry-picking anymore. So, in v6: - Converted the NMI tty driver to use tty_port helpers, per Alan Cox's suggestions; - In uart's poll_init callback fixed a race, spotted by Alan; - Use test_bit instead of touching port->flags directly; These patches can be found in the following repo (based on tty-next): git://git.infradead.org/users/cbou/linux-nmi-kdb.git master Old changelogs and rationale for these patches can be found here: v1-v5: http://lkml.org/lkml/2012/9/10/2 Thanks, -- arch/arm/Kconfig | 19 ++ arch/arm/common/vic.c | 28 +++ arch/arm/include/asm/hardware/vic.h | 2 + arch/arm/include/asm/kgdb.h | 8 + arch/arm/kernel/Makefile | 1 + arch/arm/kernel/entry-armv.S | 167 +------------ arch/arm/kernel/entry-header.S | 170 +++++++++++++ arch/arm/kernel/kgdb_fiq.c | 99 ++++++++ arch/arm/kernel/kgdb_fiq_entry.S | 87 +++++++ arch/arm/mach-versatile/Makefile | 1 + arch/arm/mach-versatile/kgdb_fiq.c | 31 +++ drivers/tty/serial/Kconfig | 19 ++ drivers/tty/serial/Makefile | 1 + drivers/tty/serial/amba-pl011.c | 66 ++++- drivers/tty/serial/kgdb_nmi.c | 391 ++++++++++++++++++++++++++++++ drivers/tty/serial/kgdboc.c | 16 ++ drivers/tty/serial/serial_core.c | 32 +++ include/linux/kdb.h | 29 ++- include/linux/kgdb.h | 34 +++ include/linux/serial_core.h | 2 + include/linux/tty_driver.h | 1 + kernel/debug/debug_core.c | 36 ++- kernel/debug/kdb/kdb_main.c | 29 +++ 23 files changed, 1076 insertions(+), 193 deletions(-)

12 years, 8 months

4
22
0 0

[PATCH v5 0/14] KGDB/KDB FIQ (NMI) debugger

by Anton Vorontsov

Hi all, There wasn't much feedback on v4, the only comment was from Brian Swetland concerning async console (I explained how we deal with it). It would be really great if the core functionality could make it into v3.7. Which raises the question: if the patches are OK, who should take them? They touch 3 subsystems: KGDB, TTY and ARM. Taking the patches via -mm or TTY trees would be the easiest way as this way we'll avoid having to deal with conflicts (see changelog*). But merging via ARM or KDB will also work. Russell, Jason, naively presuming that the ARM & KDB patches are OK, would you be willing to ack ARM/KDB patches? Or, in case if it goes via KDB or ARM tree, we'll need acks from Greg and Alan on tty patches... Anyways, here goes the shiny v5: - *I took two amba-pl1011 patches from Greg's tty tree. This is needed to ease Stephen Rothwell's life in case if this goes into -next via non-tty or non-mm tree. The problem is that we now touch the same lines as tty tree, and conflicts are not trivial. But by taking the two patches and rebasing my work on top, we turn the conflicts into trivial ones. - There were some concerns that '$3#33' might be not lengthy enough (i.e., it's a bit shorter than '\nreboot\n'). Reading 2GB of /dev/urandom couldn't find $3#33 sequence, but I made the magic phrase configurable via kgdb_nmi.magic kernel command line option, just in case. These patches can be found in the following repo: git://git.infradead.org/users/cbou/linux-nmi-kdb.git master Boilerplate: These patches introduce KGDB FIQ debugger support. The idea (and some code, of course) comes from Google's FIQ debugger[2]. There are some differences (mostly implementation details, feature-wise they're almost equivalent, or can be made equivalent, if desired). The FIQ debugger is a facility that can be used to debug situations when the kernel stuck in uninterruptable sections, e.g. the kernel infinitely loops or deadlocked in an interrupt or with interrupts disabled. On some development boards there is even a special NMI button, which is very useful for debugging weird kernel hangs. And FIQ is basically an NMI, it has a higher priority than IRQs, and upon IRQ exception FIQs are not disabled. It is still possible to disable FIQs (as well as some "NMIs" on other architectures), but via special means. So, here FIQs and NMIs are synonyms, but in the code I use NMI term for arch-independent code, and FIQs for ARM code. A few years ago KDB wasn't yet ready for production, or even not well-known, so originally Google implemented its own FIQ debugger that included its own shell, ring-buffer, commands, dumping, backtracing logic and whatnot. This is very much like PowerPC's xmon (arch/powerpc/xmon), except that xmon was there for a decade, so it even predates KDB. Anyway, nowadays KGDB/KDB is the cross-platform debugger, and the only feature that was missing is NMI handling. This is now fixed for ARM. There are a few differences comparing to the original (Google's) FIQ debugger: - Doing stuff in FIQ context is dangerous, as there we are not allowed to cause aborts or faults. In the original FIQ debugger there was a "signal" software-induced interrupt, upon exit from FIQ it would fire, and we would continue to execute "dangerous" commands from there. In KGDB/KDB we don't use signal interrupts. We can do easier: set up a breakpoint, continue, and you'll trap into KGDB again in a safe context. It works for most cases, but I can imagine cases when you can't set up a breakpoint. For these cases we'd better introduce a KDB command "exit_nmi", that will rise the SW IRQ, after which we're allowed to do anything. - KGDB/KDB FIQ debugger shell is synchronous. In Google's version you could have a dedicated shell always running in the FIQ context, so when you type something on a serial line, you won't actually cause any debugging actions, FIQ would save the characters in its own buffer and continue execution normally. But when you hit return key after the command, then the command is executed. In KGDB/KDB FIQ debugger it is different. Once you enter KGDB, the kernel will stop until you instruct it to continue. This might look as a drastic change, but it is not. There is actually no difference whether you have sync or async shell, or at least I couldn't find any use-case where this would matter at all. Anyways, it is still possible to do async shell in KDB, just don't see any need for this. - Original FIQ debugger used a custom FIQ vector handling code, w/ a lot of logic in it. In this approach I'm using the fact that FIQs are basically IRQs, except that we there are a bit more registers banked, and we can actually trap from the IRQ context. But this all does not prevent us from using a simple jump-table based approach as used in the generic ARM entry code. So, here I just reuse the generic approach. Note that I test the code on a modelled ARM machine (QEMU Versatile), so there might be some issues on a real HW, but it works in QEMU tho. :-) Assuming you have QEMU >= 1.1.0, you can easily play with the code using ARM/versatile defconfig and command like this: qemu-system-arm -nographic -machine versatilepb -kernel linux/arch/arm/boot/zImage -append "console=ttyAMA0 kgdboc=ttyAMA0 kgdb_fiq.enable=1" Thanks, -- arch/arm/Kconfig | 19 ++ arch/arm/common/vic.c | 28 +++ arch/arm/include/asm/hardware/vic.h | 2 + arch/arm/include/asm/kgdb.h | 8 + arch/arm/kernel/Makefile | 1 + arch/arm/kernel/entry-armv.S | 167 +------------- arch/arm/kernel/entry-header.S | 170 +++++++++++++++ arch/arm/kernel/kgdb_fiq.c | 99 +++++++++ arch/arm/kernel/kgdb_fiq_entry.S | 87 ++++++++ arch/arm/mach-versatile/Makefile | 1 + arch/arm/mach-versatile/kgdb_fiq.c | 31 +++ drivers/tty/serial/Kconfig | 19 ++ drivers/tty/serial/Makefile | 1 + drivers/tty/serial/amba-pl010.c | 15 +- drivers/tty/serial/amba-pl011.c | 106 +++++---- drivers/tty/serial/kgdb_nmi.c | 352 ++++++++++++++++++++++++++++++ drivers/tty/serial/kgdboc.c | 16 ++ drivers/tty/serial/serial_core.c | 30 +++ include/linux/kdb.h | 29 +-- include/linux/kgdb.h | 34 +++ include/linux/serial_core.h | 2 + include/linux/tty_driver.h | 1 + kernel/debug/debug_core.c | 36 ++- kernel/debug/kdb/kdb_main.c | 29 +++ 24 files changed, 1051 insertions(+), 232 deletions(-) In v4: - Implemented kgdb_nmi serial driver, it provides 'nmi_console' KDB command. With the driver we can use our debugger port as a normal console, except that we can always get back to the debugger using the magic sequence. Note that I still somewhat reluctant to introduce software-raised interrupts, as they're arch-specific and not always possible. So today the driver uses a tasklet, it should be pretty cheap: we're checking for the input on timer interrupts, but we don't cause needless wakeups. The pro of this is that it works everywhere (but arches still have an option to optimize things, of course); - Two new patches added to propagate init_poll() callbacks from tty to uart drivers. As a side-effect, a long-standing bug fixed in amba-pl1011 driver; - Dropped patch 'Get rid of .LCcralign local label'; - Some more fixes in SVC return path. Now it seems rock-solid; In v3: - Per Colin Cross suggestion, added a way to release a debug console for normal use. This is done via 'disable_nmi' command (in the original FIQ debugger it was 'console' command). For this I added a new callback in the tty ops, and serial drivers have to provide a way to clear its interrupts. The patch 'tty/serial/kgdboc: Add and wire up clear_irqs callback' explains the concept in details. - Made the debug entry prompt more shell-like; - A new knocking mode '-1'. It disables the feature altogether, and thus makes it possible to hook KDB entry to a dedicated button. - The code was rebased on 'v3.5 + kdb kiosk'[1] patches; In v2: - Per Colin Cross' suggestion, we should not enter the debugger on any received byte (this might be a problem when there's a noise on the serial line). So there is now an additional patch that implements "knocking" to the KDB (either via $3#33 command or return key, this is configurable); - Reworked {enable,select}_fiq/is_fiq callbacks, now multi-mach kernels should not be a problem; - For versatile machines there are run-time checks for proper UART port (kernel will scream aloud if out of range port is specified); - Added some __init annotations; - Since not every architecture defines FIQ_START, we can't just blindly select CONFIG_FIQ symbol. So ARCH_MIGHT_HAVE_FIQ introduced; - Add !THUMB2_KERNEL dependency for KGDB_FIQ, we don't support Thumb2 kernels; - New patch that is used to get rid of LCcralign label in alignment_trap macro. [1] https://lkml.org/lkml/2012/7/26/260 [2] Original Google's FIQ debugger, fiq_* files: http://android.git.linaro.org/gitweb?p=kernel/common.git;a=tree;f=arch/arm/… And board support as an example of using it: http://nv-tegra.nvidia.com/gitweb/?p=linux-2.6.git;a=commitdiff;h=461cb80c1…

12 years, 9 months

2
25
0 0

[ACTIVITY] (Appala Bade)

by Appala Bade

=== Highlights === Get LTP-DDT Test Working in LAVA: 1.Investigation on LTP-DDT, how to build it, run it, configure it : Completed 2.Get ltp-ddt running successfully on a snowball locally(Limited Testcases): Completed 3.Create a wrapper for LTP-DDT in lava-test: Started === Plans === Add ltp-ddt to appropriate kernel and/or daily runs as needed. Documentation for Driver test cases. Thanks, Appala

12 years, 9 months

1
0
0 0

[ACTIVITY] (Deepak Saxena) Sept 2 - 9 2012

by Deepak Saxena

== Deepak Saxena <dsaxena> == === Highlights === * Returned from two weeks of vacation, so not much to report here === Plans === * Catch up with all email * Brainstorm with Jakub on kernel project management * Roadmap card review with OPSCOM * Start planning for Connect * Usual 1:1's/meetings === Issues === === Travel/Time Off === * Travelling to NYC for 1 week, Sept 22nd - 28th, will be working half/three-fourth time. (This may get cancelled) * Cambridge, UK first week of October for AArch64 discussion/bring up * Copenhagen Connect

12 years, 9 months

1
0
0 0

[Activity] (Omar Ramirez) Aug 6 - Aug 12 2012

by Omar Ramirez Luna

=== omarrmz === * ARMv8t - Compile aarch64 binutils. DONE - Need to recompile the toolchain using the aarch64 binutils. DONE - Compile the aarch64 kernel with aarch64 toolchain: DONE. - Plain config generated through menuconfig. * Mailbox runtime OFF mode: DONE. - Tested on OMAP3 and OMAP4 - Needs to be submitted to LO. * Consolidate iommu changes into original series: IN PROGRESS. * Get some time to do an omap mailbox cleanup: NO UPDATE. - Low priority. * Review patches from Laurent on tidspbridge: IN PROGRESS. - Reviewed 4 fo 13 patches so far. - Review a separate bug he is reporting. === Plans === * Continue to ping Paul for my reset patches in the mailing list. * Keep on looking at aarch64 patches. Regards, Omar

12 years, 9 months

1
1
0 0

[ACTIVITY] (Shawn Guo) Week of Sept 3

by Shawn Guo

=== Highlights === * Sent v4 generic cpufreq driver cpufreq-cpu0. Rafael queued it for 3.7 for linux-pm tree. * Running into a sleep-in-atomic warning with cpufreq-cpu0 driver. It turns out that smp_twd driver calls clk_get_rate which holds a mutex in interrupt context. Dug and sent out a patch from Mike which could be a fix for review. * Mike applied patch of_clk_src_onecell_get() which is required by imx/mxs DT clk lookup. * Sent a patch on kernel/sys.c to call disable_nonboot_cpus in kernel_restart, which can help one reboot issue seen on imx6q. Andrew had it applied on -mm tree. * Reviewed patch series "Add device tree support for on-chip SRAM" from Philipp Zabel. * Reviewed a series adding device tree support for imx pwm driver from Sascha. * Reviewed a series from Dong, adding syscon driver based on regmap for general registers access. -- Regards, Shawn

12 years, 9 months

1
0
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

linaro-kernel