May 2019 - kernelnewbies@kernelnewbies.org - Kernelnewbies Mailing Lists

ULIMIT: limiting user virtual address space (stack inluded)
by Karaoui mohamed lamine May 2, 2019

May 2, 2019

Hi all, According to the man page of "ulimit", it is possible to limit the virtual address space of a user process: https://ss64.com/bash/ulimit.html To limit the virtual address space to 2 GB: $ ulimit -SH -v 1048576 However, it seems that "stack" allocation ignores this limit (I tried with both kernel 4.4 and 4.15). I found this article (patch) that seems to fix the problem: https://lwn.net/Articles/373701/ My questions are: - Why is the stack allocation not placed within the "ulimit" limit? (is it a bug?) - Is there a way, in user space, to force the stack to be allocated at a certain address? - Will the patch above (or similar) ever be integrated into the Linux kernel? Regards, Mohamed

3 2

Re: debugfs question ...
by 'Greg KH' May 1, 2019

May 1, 2019

On Wed, May 01, 2019 at 03:18:51PM +0100, rdq(a)metamail.co wrote: > Hello and thanks, > > > > > > > The pattern for the implementation is (AFAICT) right out of the book > > > > > > > You are returning a "short" read, and then disallowing ppos to be set to a > > non-zero value? That's a recipie for disaster :( > > > > Also, you allow userspace to allocate as much memory as it asks for? > > Not good :( > > Well I started here: > https://github.com/torvalds/linux/blob/v4.14/sound/soc/soc-core.c#L251. > > > Why not just use the simple_read_from_buffer() call? That handles all of > the > > "housekeeping" for you, and your function can be _much_ simpler. > > Possibly but it does not explain why returning 0 means cat displays nothing You returned nothing, so what is cat supposed to do? It worked properly. > and returning the string length causes it to loop indefinitely. You have to give userspace what it asks for, don't ignore the ppos variable. Have you tried writing a userspace program that does a correct read() call on a file or device node like this? You have to keep reading until you get all of the data you asked for, or the kernel returns end-of-file, or an error happens. You don't just do a single read and hope for the best :) > My example now reduced to: > > int ret; > char buf[64]; > int n = snprintf(buf,64,"Hello World\n"); > if (n < 0) > return -EINVAL; > // zero on success > ret = copy_to_user(user_buf,buf,n); > *ppos += n; > return ret; > > Is this the canonical form? Not at all, that's a recipe for disaster. > Using a 4.14 kernel, cat displays nothing. Or has something changed? > If so mighty odd as there are ancient example as well as the new: > > https://sites.google.com/site/linuxkernel88/sample-code/writing-a-character… in char_device_read() A char device is MUCH different from a file interface that you are using here. It kind of looks the same, but it has some subtle ways in which things can go sideways very easily. Also, that code is really really wrong, if that char_device_read() function in the above link was ever submitted to me, I would have a field day with it. Nothing like bad random internet examples to steer people in the wrong direction :( > https://github.com/torvalds/linux/blob/master/drivers/base/regmap/regmap-de… That one is not "quite" like your example here either. You have the option to handle all of the fun corner cases and error values by writing it yourself, or just use the helper function I pointed you at which does all of what you need to do, correctly. To do something in the middle is to guarantee that something will not work properly with some sort of userspace program, or be totally insecure. thanks, greg k-h

1 0

debugfs question ...
by rdq＠metamail.co May 1, 2019

May 1, 2019

Greetings, My I2C sensor driver has a debugfs entry for development purposes. Everything works fine with the exception of the read operation. When 'cat' is used, the read operation is called repeatedly and indefinitely. If the read() is changed to return 0 then, as expected, nothing is displayed. The pattern for the implementation is (AFAICT) right out of the book (shown below). What am I missing? Any thoughts much appreciated. TAIA. RDQ static ssize_t sc031gs_reg_read_file(struct file *file, char __user *user_buf, size_t count, loff_t *ppos) { char *buf = 0; ssize_t total = 0; struct sc031gs_dev *sensor = file->private_data; if (!sensor) return -EINVAL; if (*ppos < 0 || !count) return -EINVAL; buf = kmalloc(count, GFP_KERNEL); if (!buf) return -ENOMEM; total = snprintf(buf,count,"Hello world\n"); if (total >= 0) { if (copy_to_user(user_buf, buf, total)) { kfree(buf); return -EFAULT; } *ppos += total; } kfree(buf); return total; }

3 3