**** BEGIN LOGGING AT Sat Jun 11 02:59:58 2016
Jun 11 12:34:50 <kiran4399> jic23: there?
Jun 11 12:35:17 <kiran4399> jic23: I think I will do the ak8975 myself..
Jun 11 12:35:54 <kiran4399> jic23: have you build a kernel module by using the android mpu driver?? I have a strange feeling that it will work if you cross compile it..
Jun 11 12:36:07 <kiran4399> what do you think?
Jun 11 14:28:00 <ZeekHuge> Hi !
Jun 11 14:28:13 <ZeekHuge> what the benefit of using kfifo ?
Jun 11 14:28:34 <ZeekHuge> why cant we just use copy_to_user()
Jun 11 14:28:46 <ZeekHuge> or copy_from_user
Jun 11 14:29:03 <ZeekHuge> adn place it in a buffer we have in our driver
Jun 11 14:30:34 <ZeekHuge> like for example a simple char buff in our read() ? and not using a kfifo ?
Jun 11 14:31:20 <ZeekHuge> and even when we are allowing file to be opened only by one application at a time ?
Jun 11 14:32:27 <ZeekHuge> nerdboy: ds2 Abhishek_ ^
Jun 11 14:33:06 <ZeekHuge> and should a kfifo be used for in-kernel communication too ?
Jun 11 14:55:54 <ZeekHuge> okay i guess its used when the device is expected to be slower than the expected input .. right ?
Jun 11 14:58:52 <ZeekHuge> and that is probably why we use it only in read functions.
Jun 11 15:21:01 <m_w> I am here now
Jun 11 15:23:45 <ZeekHuge> hey m_w :)
Jun 11 15:23:56 <m_w> whats up?
Jun 11 15:24:14 <ZeekHuge> I am fine . I got y driver up and speaking to pru
Jun 11 15:24:19 <ZeekHuge> *my
Jun 11 15:25:00 <m_w> excellent
Jun 11 15:25:02 <ZeekHuge> and had some simple doubts .. should repost ? its just the last part in the logs ?
Jun 11 15:25:12 <ZeekHuge>  what the benefit of using kfifo ?
Jun 11 15:25:12 <ZeekHuge> <ZeekHuge> why cant we just use copy_to_user()
Jun 11 15:25:12 <ZeekHuge> <ZeekHuge> or copy_from_user
Jun 11 15:25:12 <ZeekHuge> <ZeekHuge> adn place it in a buffer we have in our driver
Jun 11 15:25:12 <ZeekHuge> <ZeekHuge> like for example a simple char buff in our read() ? and not using a kfifo ?
Jun 11 15:25:12 <ZeekHuge> <ZeekHuge> and even when we are allowing file to be opened only by one application at a time ?
Jun 11 15:25:14 <ZeekHuge> <ZeekHuge> nerdboy: ds2 Abhishek_ ^
Jun 11 15:25:15 <m_w> all nighter paid off a bit
Jun 11 15:25:16 <ZeekHuge> <ZeekHuge> and should a kfifo be used for in-kernel communication too ?
Jun 11 15:25:18 <ZeekHuge> <ZeekHuge> okay i guess its used when the device is expected to be slower than the expected input .. right ?
Jun 11 15:25:21 <ZeekHuge> <ZeekHuge> and that is probably why we use it only in read functions.
Jun 11 15:25:59 <ZeekHuge> yeah. Actually, I fell asleep at 3 or maybe 4 am. And then, it all is really confusing.
Jun 11 15:26:52 <ZeekHuge> I will write a post, but once I gain enough confidence that i understand all this.
Jun 11 15:28:34 <ZeekHuge> m_w: ^ ?
Jun 11 15:28:45 <ZeekHuge> is that correct ? about kfifo
Jun 11 15:28:46 <ZeekHuge> ?
Jun 11 15:29:19 <m_w> trying to get my head around what you are asking
Jun 11 15:30:45 <m_w> so you want to put data to transfer from the PRU into a kfifo?
Jun 11 15:31:14 <ZeekHuge> nope, from kfifo to pru.
Jun 11 15:31:26 <ZeekHuge> and from the /dev file into fifo
Jun 11 15:31:31 <ZeekHuge> *kfifo
Jun 11 15:31:41 <m_w> what is going to the PRU?
Jun 11 15:31:58 <ZeekHuge> data from the cdev file.
Jun 11 15:32:10 <m_w> what data?
Jun 11 15:32:37 <ZeekHuge> any data, I am just testing the working. like, sending a number 'n'.
Jun 11 15:32:53 <m_w> okay
Jun 11 15:33:03 <ZeekHuge> and getting the leds blink n times as an output.
Jun 11 15:33:32 <m_w> sure okay
Jun 11 15:33:45 <ZeekHuge> So I saw that some driver were using kfifo to take data from cdev file, while other were using simple char array.
Jun 11 15:34:00 <ZeekHuge> my question was what to use ?
Jun 11 15:34:17 <m_w> depends on how much data is passed and how it is structured
Jun 11 15:34:46 <ZeekHuge> so, if I am expecting large input and comparatively slower comm to the device, I should use kfifo right ?
Jun 11 15:35:13 <m_w> yeah
Jun 11 15:35:32 <ZeekHuge> and I that is true for in-kernel communication as well ?
Jun 11 15:35:47 <m_w> the fifo is needed in case the driver is flooded with information and can process it immediately
Jun 11 15:36:37 <m_w> can't process it immediately
Jun 11 15:37:15 <ZeekHuge> yeah okay, So even in case of in-kernel communication ? like if one driver sends data to another driver through and API ?
Jun 11 15:37:29 <ZeekHuge> *so thats true even in case
Jun 11 15:37:31 <m_w> pretty much
Jun 11 15:37:43 <ZeekHuge> okay.
Jun 11 15:37:49 <ZeekHuge> Thank you :)
Jun 11 15:38:29 <m_w> if you want to send configuration information from the userspace to the driver you should probably use ioctl
Jun 11 15:39:42 <m_w> data can be passed using read/write and copy_to/from_user
Jun 11 15:41:20 <ZeekHuge> yeah, okay.
Jun 11 15:41:33 <ZeekHuge> I am not focusing much on user interaction.
Jun 11 15:41:49 <ZeekHuge> in our case, we just need to reach upto IIO subsystem
Jun 11 15:42:27 <m_w> something like that
Jun 11 15:47:07 <m_w> pmezydlo: you here?
Jun 11 15:47:19 <pmezydlo> yes i am, hi
Jun 11 15:47:24 <pmezydlo> m_w:^
Jun 11 15:48:33 <pmezydlo> I'm trying to do something with pm_runtime
Jun 11 15:49:27 <m_w> okay
Jun 11 15:50:18 <m_w> the pm_runtime is disabling the clock after a timeout and it is woke by a sync
Jun 11 15:50:30 <pmezydlo> I am looking who it is done in a spi master which use interrupt
Jun 11 15:51:13 <m_w> well with the master, the driver controls when the transaction happens so it can run the sync when necessary
Jun 11 15:51:23 <m_w> for a slave this is not the case
Jun 11 15:51:58 <m_w> the slave must handle the transactions asyncronously and cannot be sure when it is going to happen
Jun 11 15:54:17 <pmezydlo> yeah I know, when interrupt is caused We must sync again
Jun 11 15:54:51 <pmezydlo> This will not be the optimal
Jun 11 15:56:34 <pmezydlo> m_w: no terminal after reset is clear
Jun 11 15:58:11 <m_w> perhaps we can manage the clock directly
Jun 11 15:58:33 <m_w> just enable it in the probe and disable it in the remove
Jun 11 15:59:03 <pmezydlo> Yes, exactly
Jun 11 15:59:22 <pmezydlo> In the master driver it works
Jun 11 16:01:54 <m_w> try setting the delay negative with pm_runtime_set_autosuspend_delay
Jun 11 16:02:21 <m_w> this should prevent autosuspend
Jun 11 16:03:30 <m_w> #define SPI_AUTOSUSPEND_TIMEOUT -1
Jun 11 16:06:29 <kiran4399> m_w: did you anytime try compiling the android invensense drivers on beaglebone black?
Jun 11 16:07:41 <m_w> kiran4399: no I have only dealt with the mainline driver
Jun 11 16:07:57 <pmezydlo> m_w: driver again reset bbb
Jun 11 16:08:44 <pmezydlo> m_w: I will commit the current code on github
Jun 11 16:08:57 <m_w> okay
Jun 11 16:09:05 <m_w> something strange is going on
Jun 11 16:10:10 <pmezydlo> when spi slave driver isn't intall master receiver only 0 when slave is install master receive  FF
Jun 11 16:11:07 <pmezydlo> but slave board in this moment is reset
Jun 11 16:11:49 <m_w> so as soon as the SCLK toggles?
Jun 11 16:12:12 <m_w> you can try toggling it manually and see if it causes the reset
Jun 11 16:13:17 <pmezydlo> https://github.com/pmezydlo/SPI_slave_driver_implementation/blob/master/driver/spi-mcspi-slave.c
Jun 11 16:14:09 <pmezydlo> ok I will try
Jun 11 16:14:33 <m_w> https://github.com/pmezydlo/SPI_slave_driver_implementation/blob/master/driver/spi-mcspi-slave.c#L269
Jun 11 16:14:36 <m_w> what is this?
Jun 11 16:15:55 <m_w> take that out
Jun 11 16:16:53 <pmezydlo> I started writing support for interrupts. This is trick would the compiler does not give warnings
Jun 11 16:17:22 <m_w> oh
Jun 11 16:17:51 <m_w> perhaps you should take the pr_info out of the interrupt function
Jun 11 16:18:03 <m_w> and see if that helps
Jun 11 16:19:07 <m_w> meanwhile I will power up my BBB
Jun 11 16:21:35 <pmezydlo> m_w:still reset
Jun 11 16:22:15 <m_w> so as soon as the sclk toggles
Jun 11 16:23:02 <pmezydlo> simply remove the SCLK pin
Jun 11 16:25:57 <m_w> :-/
Jun 11 16:29:07 <m_w> I really need a USB to serial ttl adapter for this board
Jun 11 16:39:30 <m_w> building the driver now
Jun 11 16:40:56 <pmezydlo1> travis not built, I'm checking log
Jun 11 16:41:23 <pmezydlo1> kernel coding style error
Jun 11 16:43:07 <m_w> :D
Jun 11 16:43:09 <m_w> so
Jun 11 16:43:28 <m_w> I get a NULL pointer dereference when I load the dtbo
Jun 11 16:44:22 <m_w> echo SPI0_slave > /sys/devices/platform/bone_capemgr/slots
Jun 11 16:44:36 <m_w> blaaaaaaaaah
Jun 11 16:44:50 <pmezydlo1> haha
Jun 11 16:46:13 <m_w> maybe I missed a step
Jun 11 16:46:34 <m_w> like blacklisting the other module
Jun 11 16:53:31 <m_w> okay it loaded
Jun 11 16:58:37 <m_w> I am toggling the clock and it is not resetting
Jun 11 16:59:57 <m_w> oops wrong channel
Jun 11 17:00:42 <m_w> yup it is resetting
Jun 11 17:01:43 <pmezydlo1> ohh no
Jun 11 17:02:00 <m_w> it actually seemed to halt mine
Jun 11 17:02:11 <m_w> the led stopped blinking
Jun 11 17:06:17 <m_w> the driver cries a little when you rmmod it
Jun 11 17:06:40 <m_w> when you use devm_request_irq you don't explicitly free it
Jun 11 17:06:48 <m_w> it happens automatically
Jun 11 17:07:16 <pmezydlo1> fine I don't know
Jun 11 17:07:33 <pmezydlo1> What do you think , What's going on?
Jun 11 17:08:01 <kiran4399> Does anybody know how to compile a kernel module for beaglebone black?
Jun 11 17:08:58 <m_w> :D
Jun 11 17:09:20 <m_w> we are compiling a module right now
Jun 11 17:09:44 <m_w> I am a bit confused as to what is happening
Jun 11 17:12:00 <pmezydlo1> This is a hardware problem?
Jun 11 17:12:35 <m_w> I hope not
Jun 11 17:15:44 <m_w> very strange
Jun 11 17:17:35 <ZeekHuge> kiran4399: make KDIR=/path/to/kernel/header/dir ARCH=arm CROSS_COMPILE=arm-linux-gnueabihf-  LOCALVERSION=-ti-r29 SUBDIRS=/to/the/module/dir modules
Jun 11 17:17:43 <ZeekHuge> ?
Jun 11 17:22:31 <kiran4399> ZeekHuge: Nothing to be done for all error?
Jun 11 17:22:43 <kiran4399> make: Nothing to be done for all
Jun 11 17:24:15 <ZeekHuge> probably check where you are (the dir ) , and KDIR and SUBDIRS .
Jun 11 17:24:26 <ZeekHuge> kiran4399: exact command ?
Jun 11 17:24:33 <ZeekHuge> I have used that many times.
Jun 11 17:25:04 <ZeekHuge> you have a make file at SUBDIRS right ?
Jun 11 17:25:24 <kiran4399> http://pastebin.com/hm3jvmYg
Jun 11 17:25:48 <kiran4399> I have this makefile in the repository where the entire source code is there..
Jun 11 17:29:04 <ZeekHuge> okay so compiling on the BBB itself.
Jun 11 17:29:11 <ZeekHuge> makefile is wrong
Jun 11 17:29:24 <ZeekHuge> newline after all:
Jun 11 17:29:27 <ZeekHuge> and clean:
Jun 11 17:30:08 <ZeekHuge> kiran4399: ^
Jun 11 17:30:41 <ZeekHuge> did that work !
Jun 11 17:30:43 <ZeekHuge> ?
Jun 11 17:30:47 <ZeekHuge> kiran4399: ^
Jun 11 17:36:53 <ZeekHuge> kiran4399: http://paste.debian.net/737836/
Jun 11 17:37:34 <pmezydlo1> m_w: ?
Jun 11 17:40:02 <m_w> you should consider looking at (MCSPI_SYSCONFIG)
Jun 11 17:58:39 <Abhishek_> ZeekHuge: copy_to_user/copy_from_user is exactly how beaglelogic does it
Jun 11 17:59:21 <Abhishek_> It has ioctl calls but also it has sysfs attributes that can be changes
Jun 11 17:59:24 <Abhishek_> *charged
Jun 11 17:59:34 <Abhishek_> *changed
Jun 11 18:01:07 <ZeekHuge> Abhishek_: okay, but in our case, we will just come up to IIO subsystem, so dont need to much worry about it . right ?
Jun 11 18:01:50 <Abhishek_> jic23 will be able to answer the specifics in this case.
Jun 11 18:02:12 <ZeekHuge> I was basically thinking, if we will have to use kfifo for in-kernel messaging . and I think, we should , isnt it ?
Jun 11 18:03:35 <ZeekHuge> Abhishek_: okay, I'll ask him about this. I was saying this with reference to page 4 https://docs.google.com/document/d/1GfnoWusWUeH2VsLbt1qJaRi20uNxCC8O_Xw1CHoGY9A/edit
Jun 11 18:03:44 <ZeekHuge> the proposal that is
Jun 11 18:06:36 <m_w> pmezydlo1: 172:          4      INTC  65 Level     spi-mcspi-slave
Jun 11 18:06:40 <m_w> ;)
Jun 11 18:07:09 <pmezydlo1> I must set SIDLEMODE register?
Jun 11 18:07:12 <pmezydlo1> ?
Jun 11 18:07:19 <pmezydlo1> 172:          4      INTC  65 Level     spi-mcspi-slave?
Jun 11 18:07:35 <m_w> you need to clear the interrupts in the interrupt service routine
Jun 11 18:08:58 <pmezydlo1> MCSPI_IRQSTATUS?
Jun 11 18:09:00 <m_w> mcspi_slave_write_reg(slave->base, MCSPI_IRQSTATUS, 0xffffffff);
Jun 11 18:09:24 <m_w> in mcspi_slave_irq
Jun 11 18:09:29 <pmezydlo1> yeah
Jun 11 18:13:27 <pmezydlo1> anything else?
Jun 11 18:14:06 <pmezydlo1> I must set in SYSCONFIG?
Jun 11 18:14:30 <pmezydlo1> m_w: ^^
Jun 11 18:14:55 <m_w> it seems not require with the setup code that you currently have
Jun 11 18:15:20 <m_w> required
Jun 11 18:16:04 <m_w> http://pastebin.com/4fCvfqd0
Jun 11 18:16:10 <pmezydlo1> thanks
Jun 11 18:16:26 <m_w> no problem
Jun 11 18:16:36 <m_w> of to the grocery store
Jun 11 18:16:37 <pmezydlo1> I see interrupt
Jun 11 18:16:38 <m_w> off
Jun 11 18:17:10 <m_w> you should be okay to continue now
Jun 11 18:17:54 <m_w> email if you run into more troubles
Jun 11 18:28:21 <jic23> Hi all, very brief visit. Kiran adding somewhat aged android driver will prob be more of a pain than the few patches to do it with mainline driver.
Jun 11 18:28:48 <jic23> Kiran4399^^
Jun 11 18:30:11 <jic23> Plus aim is to work with upstream. The android driver mess is never going to be accepted upstream as that way we need to have two copies of every driver ( 3 actually as St have a part with same rough hardware)
Jun 11 18:30:22 <jic23> Back later
Jun 11 18:49:24 <m_w> kiran4399: did you figure everything out?
Jun 11 19:00:36 * nerdboy has a decent test image build but still no sdcard_img bbclass
Jun 11 19:00:52 <chanakya_vc> hey m_w I have a doubt in a makefile. Can you help out?
Jun 11 19:01:19 <chanakya_vc> *me
Jun 11 19:01:32 <chanakya_vc> Wormo, ^^
Jun 11 19:01:45 <Wormo> hi, what's the question?
Jun 11 19:02:32 <chanakya_vc> Wormo,Please take a look at the makefile here:https://github.com/chanakya-vc/PRU-I2C_SPI_master/tree/wip_on_spi
Jun 11 19:02:36 <chanakya_vc> In the repo
Jun 11 19:03:10 <chanakya_vc> Now if you see the makefile,there is the include path field and lib field,which I have set to my local path
Jun 11 19:03:39 <chanakya_vc> line 21 and line 22
Jun 11 19:03:50 <Wormo> yes
Jun 11 19:04:14 <chanakya_vc> Now I don't want people to always open the makefile and change that path to their local path
Jun 11 19:04:44 <chanakya_vc> Wormo, Is there a way to define a variable like PRU_CGT(as they have done in the beginning)
Jun 11 19:05:03 <Wormo> the most basic improvement is to turn those into variables, yes
Jun 11 19:05:47 <chanakya_vc> So that the user simply sets that variable and they can set those paths? Can you please guide me?I know how to declare a variable in a makefile
Jun 11 19:06:08 <Wormo> LIBS and INCLUDE are variables :)
Jun 11 19:06:28 <chanakya_vc> Wormo, But how do I prompt the user to set them?
Jun 11 19:06:57 <Wormo> You would actually make another variable that is used in both of those
Jun 11 19:07:15 <Wormo> see how there is a base path in common right
Jun 11 19:07:26 <Wormo> substitute a variable value for those parts
Jun 11 19:07:38 <chanakya_vc> And the way to set them would be to say: export LIBS =<path>?
Jun 11 19:07:57 <Wormo> don't have them export LIBS directly
Jun 11 19:08:20 <Wormo> factor out the path to the PRU software, and have them set that, PRU_SW_PATH or something
Jun 11 19:08:48 <Wormo> export PRU_SW_PATH=/home/vc/Desktop/beagle-gsoc/compile/pru-software-support-package
Jun 11 19:08:53 <Wormo> and so forth
Jun 11 19:09:27 <chanakya_vc> Okay got it .Take a common variable and let it point to the PRU software pkg,because after that the path is known right?
Jun 11 19:09:27 <Wormo> then rewrite LIBS and INCLUDE in terms of that common path
Jun 11 19:09:33 <Wormo> yes
Jun 11 19:10:31 <Wormo> like how TARGET and MAP in lines 32&33 are defined by adding different suffix to common variables
Jun 11 19:11:36 <chanakya_vc> One more fundamental doubt I have,is that in linux when you do export var=<>,then does bash initialize a variable by that name?Because before the makefile the variable PRU_SW_PATH would not be defined anywhere else right?
Jun 11 19:12:25 <Wormo> It creates the variable with 'export' yes, you don't need to pre-declare like in C
Jun 11 19:12:34 <chanakya_vc> Wormo, I mean the makefile must look into the host computer for a variable by that name and then update its own value right?
Jun 11 19:12:41 <chanakya_vc> Okay
Jun 11 19:13:12 <Wormo> 'make' looks in the environment if you have not set the variable in the Makefile
Jun 11 19:13:59 <chanakya_vc> Okay. And I am modifying this TI makefile,so how do I go about any copyright stuff?
Jun 11 19:14:02 <chanakya_vc> Wormo, ^^
Jun 11 19:14:05 <Wormo> there is also a syntax to create default values in the Makefile, where it will only use the value if user has not already stated a value in the environmenet
Jun 11 19:15:00 <Wormo> What was the copyright on the original PRU source from TI?
Jun 11 19:15:16 <chanakya_vc> *Nothing is written on the makefile
Jun 11 19:15:26 <Wormo> The whole package
Jun 11 19:15:53 <Wormo> It can be in a license file for the whole thing
Jun 11 19:16:31 <Wormo> the license per file is good practice, but it's still valid if you put it in a separate file distributed with the software
Jun 11 19:17:01 <chanakya_vc> It is actually an example given in the PRU sw support pkg.
Jun 11 19:17:19 <Wormo> yes, so what is the license for the PRU sw support pkg
Jun 11 19:17:33 <Wormo> examples fall under that unless otherwise stated
Jun 11 19:19:22 <Wormo> btw GNU Makefile syntax for default values overrideable from environment: 'VAR ?= DEFAULT_VALUE'
Jun 11 19:19:33 <chanakya_vc> Wormo, There is a notice on top of the example code which says http://pastebin.com/7U37RysM
Jun 11 19:19:59 <chanakya_vc> This is on top of the firmware code given along with the  makefile
Jun 11 19:20:26 <chanakya_vc> I havenot used the firmware code,Just modified the makefile
Jun 11 19:23:41 <Wormo> Looks like a BSD style license, where they pretty much just want their notice to accompany any used code
Jun 11 19:24:55 <chanakya_vc> Paste this information on top of the makefile then Wormo ?
Jun 11 19:25:04 <Wormo> you can turn that into a Makefile style comment, saying 'this makefile based TI PRU compiler examples/Makefile, with following license:'
Jun 11 19:25:19 <chanakya_vc> And add myself as one of authors?
Jun 11 19:25:24 <Wormo> yes
Jun 11 19:25:55 <chanakya_vc> Okay I have not idea about how to write that in legal terms and stuff
Jun 11 19:25:56 <Wormo> your code can use a different license of course
Jun 11 19:26:55 <Wormo> Basically getting their notice in literally is legal
Jun 11 19:27:33 <Wormo> pointing to where the original Makefile came from is not a legal requirement but it's common sense
Jun 11 19:28:21 <chanakya_vc> Wormo, For the moment can I just write before the TI license, Adapted and modified by:Vaibhav Choudhay and then maybe write rights reserved and then just paste the TI license after that?
Jun 11 19:29:06 <Wormo> yes 'Adapted and modified by Vaibhav Choudhay' followed by TI license is fine
Jun 11 19:29:07 <chanakya_vc> I will worry about my license at the end when I am done
Jun 11 19:29:22 <chanakya_vc> Okay
Jun 11 19:29:32 <chanakya_vc> Thanks Wormo
Jun 11 19:29:38 <Wormo> welcome
Jun 11 19:38:42 * nerdboy getting too big with the CORE_IMAGE_EXTRA_INSTALL list in local.conf
Jun 11 19:38:57 <Wormo> how long :)
Jun 11 19:39:02 <nerdboy> i think i passed the point of "time to make an image recipe" a while back...
Jun 11 19:42:07 <Abhishek_> chanakya_vc: If the code you adapt gets significantly changed, you add your copyright alongside other authors
Jun 11 19:42:18 <Abhishek_> below them in the line
Jun 11 19:46:17 <nerdboy> https://archive.org/details/skce2009-06-16.mtx.miller.99764.flac16
Jun 11 19:55:50 <ZeekHuge> one question, if two different drivers call devm_kzalloc() with same dev as first argument, does it allocate different memories ? or just returns pointer to same memory location ?
Jun 11 19:56:49 <ZeekHuge> Wormo: nerdboy Abhishek_ ^
Jun 11 19:57:18 <Abhishek_> IIRC two different allocations happen
Jun 11 19:57:23 <Wormo> it's dynamic allocation, different pointer on different calls
Jun 11 19:57:47 <Wormo> the device is used to track what list of pointers should be freed if device goes away
Jun 11 19:58:16 <ZeekHuge> okay so then I read somewhere that if such global alloc is made, we should use mutex locks while changing the values in the ds
Jun 11 19:58:25 <ZeekHuge> so why that ?
Jun 11 19:58:45 <Abhishek_> You don't want another call to change something else while you are changing a value
Jun 11 19:59:01 <Abhishek_> because you might depend on some value that got changed in-between you making a change
Jun 11 19:59:44 <nerdboy> if you're really anal you can enable "prove locking correctness" in the kernel config...
Jun 11 20:00:09 <ZeekHuge> who else will come and change it ? if everyone has a different pointer ?
Jun 11 20:00:37 <nerdboy> might not be a bad idea...
Jun 11 20:01:20 <ZeekHuge> nerdboy: and what will that do ?
Jun 11 20:01:52 <Wormo> are you sure it was dev_kzalloc() that was supposed to be locked?
Jun 11 20:02:32 <Wormo> the lock debugging config option is a good idea, it keeps track of the order that locks have been acquired
Jun 11 20:02:38 <Abhishek_> It may be a data structure allocated using devm_kzalloc
Jun 11 20:02:55 <ZeekHuge> no ! it was said that when we change values at global alloc made by commands such as devm_kzalloc()
Jun 11 20:03:35 <ZeekHuge> we should use mutex_locks
Jun 11 20:04:32 <ZeekHuge> I am unable to find or even recall where i read that (did a lot of reading from various sources in last some days)
Jun 11 20:04:48 <Wormo> e.g. if you have one function gets mutex A then mutex B, and another function that accidentally tries to get B then A, then with just the wrong timing you can get a deadlock
Jun 11 20:04:52 <jic23> ZeekHuge, sounds crazy over generalised statement!
Jun 11 20:06:13 <Wormo> lock debugger notices this problem even if you were lucky to avoid deadlock by fuctions not running at about the same time
Jun 11 20:06:22 <jic23> On your stuff for now you may need a lock to ensure access from userspace to buffers does not mess with them getting filled from hardware.
Jun 11 20:06:55 <jic23> Lock debugger is cool but you shouldn't need it for a simple driver!
Jun 11 20:07:09 <jic23> Interesting to read up on though.
Jun 11 20:07:33 <ZeekHuge> okay.
Jun 11 20:08:09 <ZeekHuge> but I was just trying to understand and possibly use it as i might need that in further development.
Jun 11 20:08:12 <Wormo> IMO good when devs get in the habit of giving it a try sometimes, nerdboy has found some warnings in vendor kernels especially
Jun 11 20:08:25 <jic23> Be cynical if I were you. Write driver initially without locks then work out where they are needed later.
Jun 11 20:08:51 <jic23> Yeah, I have had the odd lockdep warning :)
Jun 11 20:08:58 <jic23> Oops :/
Jun 11 20:10:25 <jic23> Don't flood Zeekhuge with all the tools to early :) never get anything done!
Jun 11 20:10:30 <jic23> Too
Jun 11 20:10:49 <jic23> Back later maybe...
Jun 11 20:10:54 <ZeekHuge> :)
Jun 11 20:20:52 <chanakya_vc> Abhishek_, Got it. I am going to just write what Wormo suggested me to as of now
Jun 11 20:21:23 <chanakya_vc> Abhishek_,Thanks! : )
Jun 11 20:38:01 <Abhishek_> SG
Jun 11 22:47:59 <nerdboy> i'm pretty sure lockdep tests are separate from proof setting
**** ENDING LOGGING AT Sun Jun 12 02:59:58 2016