- kernelnewbies@kernelnewbies.org - Kernelnewbies Mailing Lists

Getting started with Development
by Ganesh Oct. 12, 2016

Oct. 12, 2016

Hello, I am interested in memory management unit of KVM. On KVM's website it is mentioned that, try LRU (currently FIFO) and analyse the performance. I want to do it as a final year B. Tech project ;but I have time limitation of 4-5 months. I have basic background of operating system and little bit about virtualization. I am eager to do this ;but I want to know that, 1. how much time does it take for new developer to implement thing like LRU? 2. Is it doable within 4-5 months? If it is NOT doable within 4-5 months then, Is their any project related to kernel development which I can do it within 4-5 months? I assure that, I will give my best Thanks !!! Regards, Ganesh

2 1

Custom ASOC codec driver
by Christian N Oct. 11, 2016

Oct. 11, 2016

Hi All I'm implementing a driver for TEF663x, and link (machine driver) to iMX6 (kernel 4.x) I understand ALSA and ASOC layers and architecture. I have good knowledge about both chipset (soc and codec). Nevertheless I am really confused with asoc structures, wich fields are mandatory and some links between platform, machine and codec. 1-Why teh followin controls are duplicated? When muste ia have to use each one? struct snd_soc_component_driver{ .... /* Default control and setup, added after probe() is run */ const struct snd_kcontrol_new *controls; unsigned int num_controls; const struct snd_soc_dapm_widget *dapm_widgets; unsigned int num_dapm_widgets; const struct snd_soc_dapm_route *dapm_routes; unsigned int num_dapm_routes; .... } but this struct is included in snd_soc_codec_driver, which one has his own widgets, routes and kcontrols struct snd_soc_codec_driver { ..... struct snd_soc_component_driver component_driver; /* Default control and setup, added after probe() is run */ const struct snd_kcontrol_new *controls; int num_controls; const struct snd_soc_dapm_widget *dapm_widgets; int num_dapm_widgets; const struct snd_soc_dapm_route *dapm_routes; int num_dapm_routes; .... } 2-In kernel/Documentation/sound/alsa/soc/platform.txt, states "The platform drivers only target the SoC CPU and must have no board specific code", but several CODEC drivers have struct snd_soc_platform_driver defined Is it an specifications missleading, or bad codec implementation? 3-Is there any document/guide to implement asoc driver, or ASOC is under development, and the mess should be organized I was looking for information at http://processors.wiki.ti.com/index.php/Sitara_Linux_Audio_DAC_Example kernel/Documentation/sound/alsa/soc ALSA home page (of curse writing alsa driver) http://blog.csdn.net/sepnic/article/category/778492 Several kernel source files (each manufacturer have his own way :( ) Thanks ind advance

1 0

userspace app receive SIGSEGV, but not is crash, why?
by Lucha Chen Oct. 11, 2016

Oct. 11, 2016

Dear All, userspace app receive SIGSEGV, but not crash, why? The mtk kernel log is : <4>[67330.976640] -(6)[25390:ndroid.systemui][25390:ndroid.systemui] sig 11 to [25390:ndroid.systemui] stat=R Thanks Lucha

2 1

Error reading /dev/mem
by Paddie O'Brien Oct. 10, 2016

Oct. 10, 2016

Hi, I have a system call that maps virtual to physical addresses. I have disabled CONFIG_STRICT_DEVMEM and am reading from /dev/mem to verify that the contents of the syscall-returned physical address match the contents of the virtual address. It works fine up to a point. Reading beyond byte 935321597 in /dev/mem throws a "bad address" error. Not sure it's relevant but I'm running on virtual box and specifying 1GB of RAM. Tried with 2GB of RAM and hit the same problem. Any idea why I can't read beyond the above point? Thanks.

3 4

if/else block default coding style question
by Nicholas Mc Guire Oct. 8, 2016

Oct. 8, 2016

Hi ! There are quite a few places (roughly 90) in the kernel where an if/else if/else block repeats the last "case" presumably as default e.g. drivers/net/wireless/realtek/rtlwifi/rtl8192c/dm_common.c if ((rtlpcipriv->bt_coexist.bt_service == BT_BUSY) && (rtlpcipriv->bt_coexist.bt_rssi_state & BT_RSSI_STATE_NORMAL_POWER)) { rtl_write_byte(rtlpriv, REG_GPIO_MUXCFG, 0xa0); } else if ((rtlpcipriv->bt_coexist.bt_service == BT_OTHER_ACTION) && (rtlpriv->mac80211.mode < WIRELESS_MODE_N_24G) && (rtlpcipriv->bt_coexist.bt_rssi_state & BT_RSSI_STATE_SPECIAL_LOW)) { rtl_write_byte(rtlpriv, REG_GPIO_MUXCFG, 0xa0); } else if (rtlpcipriv->bt_coexist.bt_service == BT_PAN) { rtl_write_byte(rtlpriv, REG_GPIO_MUXCFG, tmp1byte); } else { rtl_write_byte(rtlpriv, REG_GPIO_MUXCFG, tmp1byte); } with the last else if and else being identical. So the question is if this is accepted coding style (notably without a coment) or if things like this should be flagged. I personally find this irritating as (without a comment) it is hard to say if this is a trivial type -> missed case, or if this is intended as a default behavior. So - before starting to generate a series of patches - should stuff like this be flagged ? thx! hofrat

3 5

walk up pagetables on ARM
by Oscar Salvador Oct. 6, 2016

Oct. 6, 2016

Hi all, for learning purposes, I wanted to emulate "lookup_address" function for ARM platforms, but I'm having some troubles which I don't know how to get over them. I'll explain it: I start by retrieving the value of swapper_pg_dir from TTBR1 register (during head.S executions the physical address of swapper_pg_dir is being saved in this register). So far this works, I can compare it with gdb: (gdb) p swapper_pg_dir Cannot access memory at address 0xc0004000 (gdb) p init_mm.pgd $5 = (pgd_t *) 0xc0004000 (gdb) And from my code: static pgd_t *get_global_pgd (void) { pgd_t *pgd; unsigned int ttb_reg; asm volatile ( " mrc p15, 0, %0, c2, c0, 1" : "=r" (ttb_reg)); ttb_reg &= ~0x3fff; pgd = phys_to_virt (ttb_reg); return pgd; } and the output: kernel: [18604.342604] mod: get_global_pgd: 0x0 - c0004000 So far, it's ok, but now the fun begins. I'm walking the page table to get the PTE related to an X address: static pte_t *lookup_address (unsigned long addr) { pgd_t *pgd; pud_t *pud; pmd_t *pmd; pgd = get_global_pgd() + pgd_index (addr); pud = pud_offset (pgd, addr); pmd = pmd_offset (pud, addr); if (pmd == NULL || pmd_none (*pmd)) { pr_info ("%s: pmd == NULL\n", r2_devname); return NULL; } return pte_offset_kernel (pmd, addr); } After that, I'd to check if the PTE has the PRESENT flag. Since ARM processor doesn't provide those flags Linux had to workaround that by adding these flags: (f.i: http://lxr.free-electrons.com/source/arch/arm/include/asm/pgtable-2level.h#… ) I'm checking the flags with pte_present or pte_write But I think I'm getting the wrong PTE's because I've tried with some address from /proc/kallsyms, and it's saying that the page is not present, and I'm getting diferent values for different pages in the same range. For instance: for: 0xc0009244: bananapi kernel: [18604.419142] mod: pte_present: 0 bananapi kernel: [18604.428025] mod: pte_write: 0 bananapi kernel: [18604.436592] mod: pte_young: 2 bananapi kernel: [18604.445318] mod: pte_dirty: 0 and for 0xc0012904: bananapi kernel: [18883.095749] r2k: pte_present: 0 bananapi kernel: [18883.104607] r2k: pte_write: 1 bananapi kernel: [18883.113159] r2k: pte_young: 0 bananapi kernel: [18883.121670] r2k: pte_dirty: 0 It's saying that is not present, but I can read from those addresses with gdb: (gdb) x/2x 0xc0009244 0xc0009244 <vfp_flush_hwstate>: 0xe92d4010 0xe1a04000 (gdb) x/2x 0xc0012904 0xc0012904 <__readwrite_bug>: 0xe1a01000 0xe30505ec (gdb) What am I doing wrong? Thanks in advance

1 0

data structure for NT_386_LTS note entry in coredump
by 慕冬亮 Oct. 5, 2016

Oct. 5, 2016

Dear all, I have a question about data structure for NT_386_LTS note entry in coredump and where is the code to store this entry into coredump. When I analyzed coredump and corresponding assembly code, I found there is one memory addressing with gs. GS is a selector used by glibc. But its base address is stored in the kernel space which I don't know from "gdb executable core". But fortunately I find a note entry with type NT_386_LTS in my coredump. The following is the content of this note entry, >> LINUX 48 386_TLS >> index: 6, base: 0xb7fd57c0, limit: 0x000fffff, flags: 0x00000051 >> index: 7, base: 0x00000000, limit: 0x00000000, flags: 0x00000028 >> index: 8, base: 0x00000000, limit: 0x00000000, flags: 0x00000028 Q1: Is there a data structure defined in header file like /usr/include/elf.h? Q2: Where is the code to store this entry into coredump? -- My best regards to you. No System Is Safe! Dongliang Mu

1 0

unsubscribe
by 고영준 Oct. 5, 2016

Oct. 5, 2016

unsubscribe 블로그서명ohojang님의 블로그 소프트웨어 프로그램개발

1 0

Char device write repeating
by Kenneth Adam Miller Oct. 5, 2016

Oct. 5, 2016

I have a character device that I am calling write on and which is succeeding, but which is repeatedly executing. I have hard coded the return value to one, so I don't think the userland standard library is retrying but I could be wrong. Can anybody tell me why write would be re-executed by the kernel and how to fix it? I dont actually copy_from_user, I just need this in order to signal kernel land.

5 5

[PATCH] INITEST: Documentation: Explains how INITEST works.
by Amit Kumar Oct. 5, 2016

Oct. 5, 2016

Signed-off-by: Amit Kumar <free.amit.kumar(a)gmail.com> This is a test patch. Is this patch in right format? --- Documentation/initest.txt | 7 +++++++ 1 file changed, 7 insertions(+) create mode 100644 Documentation/initest.txt diff --git a/Documentation/initest.txt b/Documentation/initest.txt new file mode 100644 index 0000000..eed4124 --- /dev/null +++ b/Documentation/initest.txt @@ -0,0 +1,7 @@ +INITEST + +1) This is the same user space program /sbin/init, but for the purpose of testing kernel system calls. + +2) It will also use testing drivers to test internal kernel API. + +3) This is a test patch. -- 2.7.4

4 6