September 2016 - kernelnewbies@kernelnewbies.org

Redundant code in fn identity_mapping of intel-iommu.c (intel IOMMU driver)?
by Rawcoder Sept. 18, 2016

Sept. 18, 2016

Hi, I hope this is the right place for driver code related queries. I came across some redundant checks in function identity_mapping of file intel-iommu.c (https://github.com/torvalds/linux/blob/master/drivers/iommu/intel-iommu.c#L…) This function is called by function iommu_no_mapping of same file (https://github.com/torvalds/linux/blob/master/drivers/iommu/intel-iommu.c#L…) static int iommu_no_mapping(struct device *dev) { int found; if (iommu_dummy(dev)) /* dev->archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO */ return 1; if (!iommu_identity_mapping) return 0; found = identity_mapping(dev); /* this function */ if (found) { // ... The function iommu_dummy is checking whether the device belongs to the dummy domain, DUMMY_DEVICE_DOMAIN_INFO. Followed by that we're checking whether identity mapping is set. After that we come to the identity_mapping function (comments are mine): static int identity_mapping(struct device *dev) { struct device_domain_info *info; if (likely(!iommu_identity_mapping)) /* we just checked this expression and we are here because it was false */ return 0; info = dev->archdata.iommu; if (info && info != DUMMY_DEVICE_DOMAIN_INFO) /* this is just !iommu_dummy(dev); which we know is the case */ return (info->domain == si_domain); return 0; } If we inline that code we get: int found; struct device_domain_info *info; /* iommu_dummy */ if (dev->archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO) return 1; if (!iommu_identity_mapping) return 0; /* identity_mapping */ if (likely(!iommu_identity_mapping)) /* redundant? */ found = 0; found = 0; info = dev->archdata.iommu; if (info && info != DUMMY_DEVICE_DOMAIN_INFO) /* redundant? */ found = (info->domain == si_domain); No other function is calling identity_mapping so it couldn't be the case that there can some other flows which require these checks. I'm not sure whether this is intentional (doesn't seem to me). Can anyone confirm this? I'm a newbie so I posted here first to check whether I'm not making that "newbie" mistake. Thanks and regards, Abhishek

1 0

Eduyptula task 2 doubt.
by Vipul Jain Sept. 16, 2016

Sept. 16, 2016

Hi, I got kernel from git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git and build and booted my kernel with CONFIG_LOCALVERSION_AUTO set and uname -r comes out as 4.8.0-rc5-42-desktop and even the git tag list v4.8-rc5 as the latest but still penguin did not accept my task just wanted to know is penguin correct and I need to focus on getting the latest linux source or this is penguin issue. Please kindly let me know I have submitted twice to penguin but he does not seems to be happy. Regards, Vipul.

4 4

Problem in notifier chain for net device
by Gurpartap Singh Sept. 16, 2016

Sept. 16, 2016

static int edal_netdevice_event(struct notifier_block *this, unsigned long event, void *ptr) { struct net_device *dev = netdev_notifier_info_to_dev(ptr); // struct net_device *dev = NULL; // dev = ptr; switch (event) { case NETDEV_REGISTER: printk(KERN_INFO"###### EDAL NET Interface is Up #######\n"); printk(KERN_INFO"EDAL MESSAGE: %s IS UP\n",dev->name); break; case NETDEV_UNREGISTER: printk(KERN_INFO"###### EDAL NET Interface is Down ######\n"); printk(KERN_INFO"EDAL MESSAGE: %s IS Down\n",dev->name); break; } return 0; } static struct notifier_block edal_netdevice_notifier= { .notifier_call = edal_netdevice_event, .next = NULL, .priority = 1, }; static struct notifier_block edal_netdevice_notifier; static int __init initial(void) { printk(KERN_INFO"This is initialisation function=%s\n",__func__); printk(KERN_INFO"hello kernel\n"); register_netdevice_notifier(&edal_netdevice_notifier); return 0; } void __exit cleanup(void) { printk(KERN_INFO"This is exit function=%s\n",__func__); printk(KERN_INFO"bye kernel\n"); unregister_netdevice_notifier(&edal_netdevice_notifier); } module_init(initial); module_exit(cleanup); When I insmod the module it shows: ###### EDAL NET Interface is Up ####### [ 1315.348657] EDAL MESSAGE: lo IS UP [ 1315.348659] ###### EDAL NET Interface is Up ####### [ 1315.348661] EDAL MESSAGE: eth0 IS UP [ 1315.348663] ###### EDAL NET Interface is Up ####### [ 1315.348665] EDAL MESSAGE: wlan0 IS UP but when I remove the Ethernet Interface It doesn't show anything but when I remove module then it shows: [ 1171.403424] This is exit function=cleanup [ 1171.403428] bye kernel [ 1171.403431] ###### EDAL NET Interface is Down ###### [ 1171.403432] EDAL MESSAGE: lo IS Down [ 1171.403434] ###### EDAL NET Interface is Down ###### [ 1171.403435] EDAL MESSAGE: eth0 IS Down [ 1171.403437] ###### EDAL NET Interface is Down ###### [ 1171.403438] EDAL MESSAGE: wlan0 IS Down Please help!

1 0

Q: what is the need for console write callback if there is serial tx (start_tx) ?
by Ran Shalit Sept. 16, 2016

Sept. 16, 2016

Hello, I am writing a driver with serial tx/rx. I am not sure what is the need console routines, when there is serial/uart tx/rx. Isn't writing to console, the same as outputing chars to serial ? I've tried to find information on this, but found none. Thank you, Ran

3 3

device_create_file() and device_remove_file()
by Matias Mucciolo Sept. 14, 2016

Sept. 14, 2016

Hi im looking at the driver: drivers/staging/olpc_dcon/olpc_dcon.c (trying to clean up __ATTR with DEVICE_ATTR_RO/_RW) line 663 creates sysfs files(in a for) : ...... rc = device_create_file(&dcon_device->dev, &dcon_device_files[i]); if (rc) { dev_err(&dcon_device->dev, "Cannot create sysfs file\n"); goto ecreate; } ... if it fails, then in the goto removes the created ones with device_remove_file() and exits the question i have is: if don't fails and the files are created ... in that code i don't see any other device_remove_file() call to remove the files when, for example, the module unload. i thinking its because when its load and used by the device they can't be unload ?? so they don't need to clean up those files ? or i missing something ? Thanks -- Matias Mucciolo Area de Infraestructura. Piedras 737 C.A.B.A SUTEBA

1 0

Re: Status/Progress of CherryTrail / BayTrail support in mainline Linux kernel
by Hanno Zulla Sept. 12, 2016

Sept. 12, 2016

Hi, >> Works for me! :) > Same here. Good for you. :-) >> All you have is an issue with a non-Intel wifi chip? Good to have that clarified, as I was mistaken about that the Realtek wifi was somehow part of the Intel SOC. (I *am* a newbie, after all.) The issue was taken to linux-wireless by now and hopefully will see some progress there. >> What else is wrong? As the on-board wifi didn't work, there is now a USB wifi stick in the netbook. That stick (rtl8188cus) works fine on other devices, including RPi. But on this netbook, a wifi connection is lost for unknown reasons after a while and I have to reconnect to the network. No idea what is causing this, the logs were inconclusive on this. > I always use latest kernel. The problem is third party devices like > WiFi, touchpad etc. As GKH suggested enter bugs, if the problem is > in Intel SoC, we always address. Thanks. I'll try to report bugs to the proper channels in the future. Kind regards & keep up the good work, Hanno

1 0

Status/Progress of CherryTrail / BayTrail support in mainline Linux kernel
by Hanno Zulla Sept. 12, 2016

Sept. 12, 2016

Hi, sorry for asking a possibly obvious question, but: Why is support for Cherry Trail / Bay Trail so bad in the mainline kernel? Is this being worked on? And where can one help or track the progress with improving this? Recently, I bought a Win10 netbook to install Linux on. It's a beautiful small device with a fanless Z3735F CPU. Just right for my kids as their first own computer. After using Intel hardware since several years without a hitch, it was surprising to find this not being fully supported by current 2016 Linux distributions. The Wifi is not running at all, lost the network with a Wifi stick that works fine on RPi and the system freezes every now and then for no reproducible reasons. Googling around, all I found were similar complaints by other Linux users and a confusing array of incomplete third-party driver projects not ready for mainline, like this one: https://github.com/hadess/rtl8723bs/issues/96 Also, there's a litany of "don't buy Cherry Trail or Bay Trail devices for Linux" articles, which probably would have been helpful to read before buying that fateful device. This is sad to watch. I tried to find a different device for my kids, but all of the netbook-like devices on the market now share this problem. A whole generation of very interesting devices appears to be lost for Linux users. Is there hope for this device class or does one have to stick to Core i CPUs for the time being? Kind regards, Hanno

3 2

check if a kernel page is read-only
by Oscar Salvador Sept. 12, 2016

Sept. 12, 2016

Hi all, I'm writing a module to read/write kernel memory, and for this I'd like to check if a page is marked as read-only, so we don't have the right to write there (kernel code for instance). I though about the flags field from "struct page {}", but from the name I can't see any useful flag to check this. The only flag I saw is different between kernel code and module code is that kernel code has the flag "reserved" enabled, so it's not swapped. Could someone provide me a hint about that? Thank you very much

3 9

AW:Getting stacktrace for userspace applications on crash(SIGSEGV)
by Johannes Thoma Sept. 12, 2016

Sept. 12, 2016

2 3

udelay vs usleep_range
by Gargi Sharma Sept. 11, 2016

Sept. 11, 2016

Hi all, I ran the checkpatch script over drivers/staging/nvec.c and got the following warning udelay(33); CHECK: usleep_range is preferred over udelay; see Documentation/timers/timers-howto.txt I checked out the timers-howto.txt and for usleep_range, one has to specify the lower and upper limit for the timer. How does one decide what the upper limit will be if one is to change the function from udelay to usleep_range? Thanks! Gargi

3 4