Query regarding physically discontiguous memory
Hi, I just started reading Memory Managment chapter from book Professional Linux Kernel Architecture. At very start of chapter I came across terms FLATMEM, DISCONTIGMEM and SPARSEMEM. Out of these DISCONTIGMEM and SPARSEMEM are available to NUMA systens only. Book also says UMA systems can also choose DISCONTIGMEM option if there address space contains large holes. Some googling says DISCONTIGMEM is prefered when we have physical discontiguous memory. This is what I don't understand what does it mean by containing large holes? physical discontiguous memory? Is it like when I put 2 RAM chips in a system then memory for them will be discontiguous? Would /proc/iomem give me details I wanted? Per my understanding RAM memory will ALWAYS be contiguous whether its 512 MB or 4 GB. Please give me graceful example of discontiguous physical memory That will really help me out. ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Spinlocks schedule_timeout
Hi, I wanted to verify/clarify if using schedule_timeout after acquiring a spinlock is safe?. So basically you have a situation where you are working on a critical code and you acquire a spin lock and you end up using a schedule_timeout after say setting the current task to Interruptible state 'cause you are waiting for some user space code to fill in some buffer from which you want to read. So my doubt is simple, is it safe to acquire a spinlock and use schedule_timeout. I assume the task is put in the run queue and will ALWAYS be scheduled back after the specified delay in schedule_timeout. -- *Regards,Santosh* ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: Spinlocks schedule_timeout
Sent from my Sony Xperia™ smartphone On Dec 25, 2013 4:16 PM, Pranay Srivastava pran...@gmail.com wrote: Hi santosh I hope you understand difference between semaphores and spin locks. Ya I am completely aware of sleeping locks n spin locks If however you are stuck a situation where you can afford being preempted then you can use sem/mutex. But i think your serialization requires that you use spin locks then best is to have a guarded variable with spin lock and release the lock before you schedule. I just wanted to verify if it was possible n if yes considering preemption is disabled.. Even if you set the state preempt count would be increased when you acquire spin lock. so you Cant schedule or go to sleep after spin lock. Since schedule_timout would make sure the task remains on the run queue and intended to be scheduled back after the specificed jeffies period. It was just out of curiosity anyway .Thanks !!! Regards On Dec 25, 2013 3:36 PM, SaNtosh kuLkarni santosh.yesop...@gmail.com wrote: Hi, I wanted to verify/clarify if using schedule_timeout after acquiring a spinlock is safe?. So basically you have a situation where you are working on a critical code and you acquire a spin lock and you end up using a schedule_timeout after say setting the current task to Interruptible state 'cause you are waiting for some user space code to fill in some buffer from which you want to read. So my doubt is simple, is it safe to acquire a spinlock and use schedule_timeout. I assume the task is put in the run queue and will ALWAYS be scheduled back after the specified delay in schedule_timeout. -- Regards, Santosh ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: how to find kernel patch applied on existing build kernel
On Wed, 25 Dec 2013 11:28:35 +0800, Dong Zhu said: rpm -q kernel --changelog That won't actually list all the 10,000+ commits between kernel releases - it often won't even list all the vendor-applied patches that they've put on top of the Linus kernel. And it assumes an RPM based system As somebody else mentioned, 'git tag --contains (commit-id)' is the best bet here. pgpaO86r94l32.pgp Description: PGP signature ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: Query regarding physically discontiguous memory
On Wed, 25 Dec 2013 14:45:01 +0530, Pritam Bankar said: This is what I don't understand what does it mean by containing large holes? physical discontiguous memory? Is it like when I put 2 RAM chips in a system then memory for them will be discontiguous? Per my understanding RAM memory will ALWAYS be contiguous whether its 512 MB or 4 GB. Your understanding is wrong. That may be true on most x86-based desktops and laptops and the like, but it's often very wrong on other architectures. In fact, given that the x86 architecture is just one of like 27 archs that Linux supports, it's a good bet that there's more archs that *don't* enforce contiguous memory than do. Counting through the defconfigs to get a number is left as an exercise for the student. :) A common case for embedded systems is 256M of RAM at address 0, and 256M of RAM at address 0x8000 (or whatever the halfway mark for the architecture is). NUMA boxes are sometimes each node's RAM is at node number*4G or similar, which means lots of holes if each node is partly populated with only 2G. pgpHfEtDXxcFF.pgp Description: PGP signature ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: Unable to access FPGA device with virtual address returned by ioremap()
Hi All, I am working on bringing up the board sbc8641d, which is mostly similar in architecture with mpc8641 having powerpc chip. My board was customized to add FPGA. This FPGA was connected as external device to processor on localbus. I am able to boot the board recognizing all the devices, except external device FPGA. I have mentioned in device tree regarding chip select and parent address in processor address domain of the fpga. I have the following problems. I have problem with ioremap returned address to my device in kernel. ioremap takes physical address and returns virtual address. Using virtual address, if I write and read, there is no result. I am writing to scratch pad of the FPGA chip and reading from it the same string. When I read I do not see any value (written value). Second, when I typed command in bash free -m it shows 756MB memory available. I used mem= commandline option to tell kernel to use 4GB memory. But after booting the kernel it still shows 756MB. Why is this? Can anyone help me in directing where to look to debug this. Any help would be appreciated. Is there anyway I can debug pagetables to see what is happening at address mapping level. NOTE: this is powerpc architecture. On Wed, Dec 25, 2013 at 3:53 PM, Sri Ram Vemulpali sri.ram.gm...@gmail.comwrote: Hi All, I am working on bringing up the board sbc8641d, which is mostly similar in architecture with mpc8641 having powerpc chip. My board was customized to add FPGA. This FPGA was connected as external device to processor on localbus. I am able to boot the board recognizing all the devices, except external device FPGA. I have mentioned in device tree regarding chip select and parent address in processor address domain of the fpga. I have the following problems. I have problem with ioremap returned address to my device in kernel. ioremap takes physical address and returns virtual address. Using virtual address, if I write and read, there is no result. I am writing to scratch pad of the FPGA chip and reading from it the same string. When I read I do not see any value (written value). Second, when I typed command in bash free -m it shows 756MB memory available. I used mem= commandline option to tell kernel to use 4GB memory. But after booting the kernel it still shows 756MB. Why is this? Can anyone help me in directing where to look to debug this. Any help would be appreciated. Is there anyway I can debug pagetables to see what is happening at address mapping level. NOTE: this is powerpc architecture. -- Regards, Sri. -- Regards, Sri. ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
MPC8641 based custom board kernel Bug
Hi, I was trying to port Linux-3.12 for MPC8641 based custom designed board for evaluation purpose. I have been facing a kernel bug at mpic initialization. Is somebody have faced this kind of bugs or can give me any pointer for further steps how to solve kernel bugs will be really helpful. here is the snapshot for the bug that may be helpful to address the bug. Using MPC86xx HPCN machine description Total memory = 512MB; using 1024kB for hash table (at cff0) Linux version 3.12.0 (ashish@ashish-VirtualBox) (gcc version 4.7.2 (GCC) ) #2 We d Dec 25 16:04:36 IST 2013 Found initrd at 0xde975000:0xdfec428a bootconsole [udbg0] enabled setup_arch: bootmem mpc86xx_hpcn_setup_arch() MPC86xx HPCN board from Freescale Semiconductor arch: exit Zone ranges: DMA [mem 0x-0x1fff] Normal empty HighMem empty Movable zone start for each node Early memory node ranges node 0: [mem 0x-0x1fff] Built 1 zonelists in Zone order, mobility grouping on. Total pages: 130048 Kernel command line: root=/dev/ram0 rw rootfs console=ttyS0,115200 PID hash table entries: 2048 (order: 1, 8192 bytes) Dentry cache hash table entries: 65536 (order: 6, 262144 bytes) Inode-cache hash table entries: 32768 (order: 5, 131072 bytes) Sorting __ex_table... Memory: 424980K/524288K available (4172K kernel code, 208K rwdata, 1304K rodata, 196K init, 149K bss, 99308K reserved, 0K highmem) Kernel virtual memory layout: * 0xfffcf000..0xf000 : fixmap * 0xff80..0xffc0 : highmem PTEs * 0xff7fe000..0xff80 : early ioremap * 0xe100..0xff7fe000 : vmalloc ioremap SLUB: HWalign=32, Order=0-3, MinObjects=0, CPUs=1, Nodes=1 NR_IRQS:512 nr_irqs:512 16 [ cut here ] kernel BUG at arch/powerpc/platforms/86xx/pic.c:42! Oops: Exception in kernel mode, sig: 5 [#1] MPC86xx HPCN Modules linked in: CPU: 0 PID: 0 Comm: swapper Not tainted 3.12.0 #2 task: c05903e0 ti: c05b4000 task.ti: c05b4000 NIP: c0567438 LR: c0567430 CTR: c0567400 REGS: c05b5ee0 TRAP: 0700 Not tainted (3.12.0) MSR: 00021032 ME,IR,DR,RI CR: 2442 XER: 2000 GPR00: c0567430 c05b5f90 c05903e0 c04e4ff8 c051e588 008f 0002 GPR08: c042789c 0001 006f 2248 bebd 11a7b4e5 200c8000 GPR16: ffbe 0024 1fec56f8 1fec59a7 GPR24: 1fff97e8 4000 1ffcc6a0 c0bff080 c05c2490 c05c2628 c0585b60 NIP [c0567438] mpc86xx_init_irq+0x38/0x108 LR [c0567430] mpc86xx_init_irq+0x30/0x108 Call Trace: [c05b5f90] [c0567430] mpc86xx_init_irq+0x30/0x108 (unreliable) [c05b5fb0] [c0562784] init_IRQ+0x24/0x38 [c05b5fc0] [c055fde4] start_kernel+0x1bc/0x2ec [c05b5ff0] [3444] 0x3444 Instruction dump: 3d00c04f 3880 38a01002 38c0 38e00100 39088f8c 3860 90010024 bfa10014 4bffec35 7c690034 5529d97e 0f09 3fa0c04f 4bfff391 3860 ---[ end trace 31fd0ba7d8756001 ]--- Kernel panic - not syncing: Attempted to kill the idle task! Rebooting in 180 seconds.. Thanks Regards Ashish Khetan ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: Spinlocks schedule_timeout
On Wed, Dec 25, 2013 at 3:29 PM, SaNtosh kuLkarni santosh.yesop...@gmail.com wrote: Hi, I wanted to verify/clarify if using schedule_timeout after acquiring a spinlock is safe?. No it is not safe. schedule_timeout is sleepy function and it is wrong to sleep when holding a spin lock. So basically you have a situation where you are working on a critical code and you acquire a spin lock and you end up using a schedule_timeout after say setting the current task to Interruptible state 'cause you are waiting for some user space code to fill in some buffer from which you want to read. So my doubt is simple, is it safe to acquire a spinlock and use schedule_timeout. I assume the task is put in the run queue and will ALWAYS be scheduled back after the specified delay in schedule_timeout. -- *Regards,Santosh* ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies Thanks, Arshad ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
