I'm trying to create a large empty file on a VFAT partition by using the `dd' command in an embedded linux box:
dd if=/dev/zero of=/mnt/flash/file bs=1M count=1 seek=1023
The intention was to skip the first 1023 blocks and write only 1 block at the end of the file, which should be very quick on a native EXT3 partition, and it indeed is. However, this operation turned out to be quite slow on a VFAT partition, along with the following message:
lowmem_shrink:: nr_to_scan=128, gfp_mask=d0, other_free=6971, min_adj=16
// ... more `lowmem_shrink' messages
Another attempt was to fopen() a file on the VFAT partition and then fseek() to the end to write the data, which has also proved slow, along with the same messages from the kernel.
So basically, is there a quick way to create the file on the VFAT partition (without traversing the first 1023 blocks)?
Thanks.
Why are VFAT "skipping" writes so slow ?
Unless the VFAT filesystem driver were made to "cheat" in this respect, creating large files on FAT-type filesystems will always take a long time. The driver, to comply with FAT specification, will have to allocate all data blocks and zero-initialize them, even if you "skip" the writes. That's because of the "cluster chaining" FAT does.
The reason for that behaviour is FAT's inability to support either:
UN*X-style "holes" in files (aka "sparse files")
that's what you're creating on ext3 with your testcase - a file with no data blocks allocated to the first 1GB-1MB of it, and a single 1MB chunk of actually committed, zero-initialized blocks) at the end.
NTFS-style "valid data length" information.
On NTFS, a file can have uninitialized blocks allocated to it, but the file's metadata will keep two size fields - one for the total size of the file, another for the number of bytes actually written to it (from the beginning of the file).
Without a specification supporting either technique, the filesystem would always have to allocate and zerofill all "intermediate" data blocks if you skip a range.
Also remember that on ext3, the technique you used does not actually allocate blocks to the file (apart from the last 1MB). If you require the blocks preallocated (not just the size of the file set large), you'll have to perform a full write there as well.
How could the VFAT driver be modified to deal with this ?
At the moment, the driver uses the Linux kernel function cont_write_begin() to start even an asynchronous write to a file; this function looks like:
/*
* For moronic filesystems that do not allow holes in file.
* We may have to extend the file.
*/
int cont_write_begin(struct file *file, struct address_space *mapping,
loff_t pos, unsigned len, unsigned flags,
struct page **pagep, void **fsdata,
get_block_t *get_block, loff_t *bytes)
{
struct inode *inode = mapping->host;
unsigned blocksize = 1 << inode->i_blkbits;
unsigned zerofrom;
int err;
err = cont_expand_zero(file, mapping, pos, bytes);
if (err)
return err;
zerofrom = *bytes & ~PAGE_CACHE_MASK;
if (pos+len > *bytes && zerofrom & (blocksize-1)) {
*bytes |= (blocksize-1);
(*bytes)++;
}
return block_write_begin(mapping, pos, len, flags, pagep, get_block);
}
That is a simple strategy but also a pagecache trasher (your log messages are a consequence of the call to cont_expand_zero() which does all the work, and is not asynchronous). If the filesystem were to split the two operations - one task to do the "real" write, and another one to do the zero filling, it'd appear snappier.
The way this could be achieved while still using the default linux filesystem utility interfaces were by internally creating two "virtual" files - one for the to-be-zerofilled area, and another for the actually-to-be-written data. The real file's directory entry and FAT cluster chain would only be updated once the background task is actually complete, by linking its last cluster with the first one of the "zerofill file" and the last cluster of that one with the first one of the "actual write file". One would also want to go for a directio write to do the zerofilling, in order to avoid trashing the pagecache.
Note: While all this is technically possible for sure, the question is how worthwhile would it be to do such a change ? Who needs this operation all the time ? What would side effects be ?
The existing (simple) code is perfectly acceptable for smaller skipping writes, you won't really notice its presence if you create a 1MB file and write a single byte at the end. It'll bite you only if you go for filesizes on the order of the limits of what the FAT filesystem allows you to do.
Other options ...
In some situations, the task at hand involves two (or more) steps:
freshly format (e.g.) a SD card with FAT
put one or more big files onto it to "pre-fill" the card
(app-dependent, optional)
pre-populate the files, or
put a loopback filesystem image into them
One of the cases I've worked on we've folded the first two - i.e. modified mkdosfs to pre-allocate/ pre-create files when making the (FAT32) filesystem. That's pretty simple, when writing the FAT tables just create allocated cluster chains instead of clusters filled with the "free" marker. It's also got the advantage that the data blocks are guaranteed to be contiguous, in case your app benefits from this. And you can decide to make mkdosfs not clear the previous contents of the data blocks. If you know, for example, that one of your preparation steps involves writing the entire data anyway or doing ext3-in-file-on-FAT (pretty common thing - linux appliance, sd card for data exchange with windows app/gui), then there's no need to zero out anything / double-write (once with zeroes, once with whatever-else). If your usecase fits this (i.e. formatting the card is a useful / normal step of the "initialize it for use" process anyway) then try it out; a suitably-modified mkdosfs is part of TomTom's dosfsutils sources, see mkdosfs.c search for the -N command line option handling.
When talking about preallocation, as mentioned, there's also posix_fallocate(). Currently on Linux when using FAT, this will do essentially the same as a manual dd ..., i.e. wait for the zerofill. But the specification of the function doesn't mandate it being synchronous. The block allocation (FAT cluster chain generation) would have to be done synchronously, but the VFAT on-disk dirent size update and the data block zerofills could be backgrounded / delayed (i.e. either done at low-prio in background or only done if explicitly requested via fdsync() / sync() so that the app can e.g. alloc blocks, write the contents with non-zeroes itself ...). That's technique / design though; I'm not aware of anyone having done that kernel modification yet, if only for experimenting.
Related
I have a very specific need:
to partially replace the content of a flash and to move MTD partition boundaries.
Current map is:
u-boot 0x000000 0x040000
u-boot-env 0x040000 0x010000
kernel 0x050000 0x230000
initrd 0x280000 0x170000
scripts 0x3f0000 0x010000
filesystem 0x400000 0xbf0000
firmware 0xff0000 0x010000
While desired output is:
u-boot 0x000000 0x040000
u-boot-env 0x040000 0x010000
kernel 0x050000 0x230000
filesystem 0x280000 0xd70000
firmware 0xff0000 0x010000
This means to collapse initrd, scripts and filesystem into a single area while leaving the others alone.
Problem is this should be achieved from the running system (booted with the "old" configuration") and I should rewrite kernel and "new" filesystem before rebooting.
The system is an embedded, so I have little space for maneuver (I have a SD card, though).
Of course the rewritten kernel will have "new" configuration written in its DTB.
Problem is transition.
Note: I have seen this Question, but it is very old and it has drawback to need kernel patches, which I would like to avoid.
NOTE2: this question has been flagged for deletion because "not about programming". I beg to disagree: I need to perform said operation on ~14k devices, most of them already sold to customers, so any workable solution should involve, at the very least, scripting.
NOTE3: if absolutely necessary I can even consider (small) kernel modifications (YES, I have means to update kernel remotely).
I will leave the Accepted answer as-is, but, for anyone who happens to come here to find a solution, I want to point out that:
Recent (<4 years old) mtd-utils, coupled with 4.0+ kernel support:
Definition of a "master" device (MTD device representing the full, unpartitioned Flash). This is a kernel option.
mtd-utils has a specific mtd-part utility that can add/delete MTD partitions dynamically. NOTE: this utility woks IF (and only if) the above is defined in Kernel.
With the above utility it's possible to build multiple, possibly overlapping partitions; use with care!
I have three ideas/suggestions:
Instead of moving the partitions, can you just split the "new" filesystem image into chunks and write them to the corresponding "old" MTD partitions? This way you don't really need to change MTD partition map. After booting into the new kernel, it will see the new contiguous root filesystem. For JFFS2 filesystem, it should be fairly straightforward to do using split or dd, flash_erase and nandwrite. Something like:
# WARNING: this script assumes that it runs from tmpfs and the old root filesystem is already unmounted.
# Also, it assumes that your shell has arithmetic evaluation, which handles hex (my busybox 1.29 ash does this).
# assuming newrootfs.img is the image of new rootfs
new_rootfs_img="newrootfs.img"
mtd_initrd="/dev/mtd3"
mtd_initrd_size=0x170000
mtd_scripts="/dev/mtd4"
mtd_scripts_size=0x010000
mtd_filesystem="/dev/mtd5"
mtd_filesystem_size=0xbf0000
# prepare chunks of new filesystem image
bs="0x1000"
# note: using arithmetic evaluation $(()) to convert from hex and do the math.
# dd doesn't handle hex numbers ("dd: invalid number '0x1000'") -- $(()) works this around
dd if="${new_rootfs_img}" of="rootfs_initrd" bs=$(( bs )) count=$(( mtd_initrd_size / bs ))
dd if="${new_rootfs_img}" of="rootfs_scripts" bs=$(( bs )) count=$(( mtd_scripts_size / bs )) skip=$(( mtd_initrd_size / bs ))
dd if="${new_rootfs_img}" of="rootfs_filesystem" bs=$(( bs )) count=$(( mtd_filesystem_size / bs )) skip=$(( ( mtd_initrd_size + mtd_scripts_size ) / bs ))
# there's no going back after this point
flash_eraseall -j "${mtd_initrd}"
flash_eraseall -j "${mtd_scripts}"
flash_eraseall -j "${mtd_filesystem}"
nandwrite -p "${mtd_initrd}" rootfs_initrd
nandwrite -p "${mtd_scripts}" rootfs_scripts
nandwrite -p "${mtd_filesystem}" rootfs_filesystem
# don't forget to update the kernel too
There is kernel support for concatenating MTD devices (which is exactly what you're trying to do). I don't see an easy way to use it, but you could create a kernel module, which concatenates the desired partitions for you into a contiguous MTD device.
In order to combine the 3 MTD partitions into one to write the new filesystem, you could create a dm-linear mapping over the 3 mtdblocks, and then turn it back into an MTD device using block2mtd. (i.e. mtdblock + device mapper linear + block2mtd) But it looks very awkward and I don't know if it'll work well (for say, OOB data).
EDIT1: added a comment explaining use of $(( bs )) -- to convert from hex as dd doesn't handle hex numbers directly (neither coreutils, nor busybox dd).
AFAIK, #andrey 's answer suggestion 1 is wrong.
an mtd partition is made of a sequence of blocks, any of which could be bad or go bad anytime. this is why the simple mtd char abstraction exists: an mtd char device (not the mtdblock one) is read sequentially and skips bad blocks. nandwrite also writes sequentially and skips bad blocks.
an mtd char device sort of acts like:
a single file into which you cannot random access, from which you can only read sequentially from the beginning to the end (or to where you get bored).
a single file into which you cannot random access, to which you can only write sequentially from the beginning (or from an erase block where you previously stopped reading) all the way to the end. (that is, you can truncate and append, but you cannot write mid-file.) to write you need to previously erase all erase blocks from where you start writing to the end of the partition.
this means that the partition size is the maximum theoretical capacity, but typically the capacity will be less due to bad blocks, and can be effectively reduced every time you rewrite the partition. you can never expect to write the full size of an mtd partition.
this is were #andrey 's suggestion 1 is wrong: it breaks up the file to be written into max-sized pieces before writing each piece. but you never know beforehand how much data will fit into an mtd partition without actually writing that data.
instead, you typically need to write some data, and you pray there will be enough good blocks to fit it. if at some point there are not, the write fails and the device reached end-of-life. needless to say, the larger the fraction of a partition you need, the higher the likelihood that the write will fail (and when that happens, it typically means that the device is toast).
to actually implement something akin to suggestion 1, you need to start writing into a partition (skipping bad blocks), and when you run out of erase blocks, you continue writing into the next partition, and so on. the point being: you cannot know where the data boundaries will lay until you actually write the data and fill each partition; there is no other way.
I want to copy files from one place to another and the problem is I deal with a lot of sparse files.
Is there any (easy) way of copying sparse files without becoming huge at the destination?
My basic code:
out, err := os.Create(bricks[0] + "/" + fileName)
in, err := os.Open(event.Name)
io.Copy(out, in)
Some background theory
Note that io.Copy() pipes raw bytes – which is sort of understandable once you consider that it pipes data from an io.Reader to an io.Writer which provide Read([]byte) and Write([]byte), correspondingly.
As such, io.Copy() is able to deal with absolutely any source providing
bytes and absolutely any sink consuming them.
On the other hand, the location of the holes in a file is a "side-channel" information which "classic" syscalls such as read(2) hide from their users.
io.Copy() is not able to convey such side-channel information in any way.
IOW, initially, file sparseness was an idea to just have efficient storage of the data behind the user's back.
So, no, there's no way io.Copy() could deal with sparse files in itself.
What to do about it
You'd need to go one level deeper and implement all this using the syscall package and some manual tinkering.
To work with holes, you should use the SEEK_HOLE and SEEK_DATA special values for the lseek(2) syscall which are, while formally non-standard, are supported by all major platforms.
Unfortunately, the support for those "whence" positions is not present
neither in the stock syscall package (as of Go 1.8.1)
nor in the golang.org/x/sys tree.
But fear not, there are two easy steps:
First, the stock syscall.Seek() is actually mapped to lseek(2)
on the relevant platforms.
Next, you'd need to figure out the correct values for SEEK_HOLE and
SEEK_DATA for the platforms you need to support.
Note that they are free to be different between different platforms!
Say, on my Linux system I can do simple
$ grep -E 'SEEK_(HOLE|DATA)' </usr/include/unistd.h
# define SEEK_DATA 3 /* Seek to next data. */
# define SEEK_HOLE 4 /* Seek to next hole. */
…to figure out the values for these symbols.
Now, say, you create a Linux-specific file in your package
containing something like
// +build linux
const (
SEEK_DATA = 3
SEEK_HOLE = 4
)
and then use these values with the syscall.Seek().
The file descriptor to pass to syscall.Seek() and friends
can be obtained from an opened file using the Fd() method
of os.File values.
The pattern to use when reading is to detect regions containing data, and read the data from them – see this for one example.
Note that this deals with reading sparse files; but if you'd want to actually transfer them as sparse – that is, with keeping this property of them, – the situation is more complicated: it appears to be even less portable, so some research and experimentation is due.
On Linux, it appears you could try to use fallocate(2) with
FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE to try to punch a hole at the
end of the file you're writing to; if that legitimately fails
(with syscall.EOPNOTSUPP), you just shovel as many zeroed blocks to the destination file as covered by the hole you're reading – in the hope
the OS will do the right thing and will convert them to a hole by itself.
Note that some filesystems do not support holes at all – as a concept.
One example is the filesystems in the FAT family.
What I'm leading you to is that inability of creating a sparse file might
actually be a property of the target filesystem in your case.
You might find Go issue #13548 "archive/tar: add support for writing tar containing sparse files" to be of interest.
One more note: you might also consider checking whether the destination directory to copy a source file resides in the same filesystem as the source file, and if this holds true, use the syscall.Rename() (on POSIX systems)
or os.Rename() to just move the file across different directories w/o
actually copying its data.
You don't need to resort to syscalls.
package main
import "os"
func main() {
f, _ := os.Create("/tmp/sparse.dat")
f.Write([]byte("start"))
f.Seek(1024*1024*10, 0)
f.Write([]byte("end"))
}
Then you'll see:
$ ls -l /tmp/sparse.dat
-rw-rw-r-- 1 soren soren 10485763 Jun 25 14:29 /tmp/sparse.dat
$ du /tmp/sparse.dat
8 /tmp/sparse.dat
It's true you can't use io.Copy as is. Instead you need to implement an alternative to io.Copy which reads a chunk from the src, checks if it's all '\0'. If it is, just dst.Seek(len(chunk), os.SEEK_CUR) to skip past that part in dst. That particular implementation is left as an exercise to the reader :)
I need to understand which files consumes iops of my hard disc. Just using "strace" will not solve my problem. I want to know, which files are really written to disc, not to page cache. I tried to use "systemtap", but I cannot understand how to find out which files (filenames or inodes) consumes my iops. Is there any tools, which will solve my problem?
Yeah, you can definitely use SystemTap for tracing that. When upper-layer (usually, a VFS subsystem) wants to issue I/O operation, it will call submit_bio and generic_make_request functions. Note that these doesn't necessary mean a single physical I/O operation. For example, writes from adjacent sectors can be merged by I/O scheduler.
The trick is how to determine file path name in generic_make_request. It is quite simple for reads, as this function will be called in the same context as read() call. Writes are usually asynchronous, so write() will simply update page cache entry and mark it as dirty, while submit_bio gets called by one of the writeback kernel threads which doesn't have info of original calling process:
Writes can be deduced by looking at page reference in bio structure -- it has mapping of struct address_space. struct file which corresponds to an open file also contains f_mapping which points to the same address_space instance and it also points to dentry containing name of the file (this can be done by using task_dentry_path)
So we would need two probes: one to capture attempts to read/write a file and save path and address_space into associative array and second to capture generic_make_request calls (this is performed by probe ioblock.request).
Here is an example script which counts IOPS:
// maps struct address_space to path name
global paths;
// IOPS per file
global iops;
// Capture attempts to read and write by VFS
probe kernel.function("vfs_read"),
kernel.function("vfs_write") {
mapping = $file->f_mapping;
// Assemble full path name for running task (task_current())
// from open file "$file" of type "struct file"
path = task_dentry_path(task_current(), $file->f_path->dentry,
$file->f_path->mnt);
paths[mapping] = path;
}
// Attach to generic_make_request()
probe ioblock.request {
for (i = 0; i < $bio->bi_vcnt ; i++) {
// Each BIO request may have more than one page
// to write
page = $bio->bi_io_vec[i]->bv_page;
mapping = #cast(page, "struct page")->mapping;
iops[paths[mapping], rw] <<< 1;
}
}
// Once per second drain iops statistics
probe timer.s(1) {
println(ctime());
foreach([path+, rw] in iops) {
printf("%3d %s %s\n", #count(iops[path, rw]),
bio_rw_str(rw), path);
}
delete iops
}
This example script is works for XFS, but needs to be updated to support AIO and volume managers (including btrfs). Plus I'm not sure how it will handle metadata reads and writes, but it is a good start ;)
If you want to know more on SystemTap you can check out my book: http://myaut.github.io/dtrace-stap-book/kernel/async.html
Maybe iotop gives you a hint about which process are doing I/O, in consequence you have an idea about the related files.
iotop --only
the --only option is used to see only processes or threads actually doing I/O, instead of showing all processes or threads
Assume we have a file of FILE_SIZE bytes, and:
FILE_SIZE <= min(page_size, physical_block_size);
file size never changes (i.e. truncate() or append write() are never performed);
file is modified only by completly overwriting its contents using:
pwrite(fd, buf, FILE_SIZE, 0);
Is it guaranteed on ext4 that:
Such writes are atomic with respect to concurrent reads?
Such writes are transactional with respect to a system crash?
(i.e., after a crash the file's contents is completely from some previous write and we'll never see a partial write or empty file)
Is the second true:
with data=ordered?
with data=journal or alternatively with journaling enabled for a single file?
(using ioctl(fd, EXT4_IOC_SETFLAGS, EXT4_JOURNAL_DATA_FL))
when physical_block_size < FILE_SIZE <= page_size?
I've found related question which links discussion from 2011. However:
I didn't find an explicit answer for my question 2.
I wonder, if the above is true, is it documented somewhere?
From my experiment it was not atomic.
Basically my experiment was to have two processes, one writer and one reader. The writer writes to a file in a loop and reader reads from the file
Writer Process:
char buf[][18] = {
"xxxxxxxxxxxxxxxx",
"yyyyyyyyyyyyyyyy"
};
i = 0;
while (1) {
pwrite(fd, buf[i], 18, 0);
i = (i + 1) % 2;
}
Reader Process
while(1) {
pread(fd, readbuf, 18, 0);
//check if readbuf is either buf[0] or buf[1]
}
After a while of running both processes, I could see that the readbuf is either xxxxxxxxxxxxxxxxyy or yyyyyyyyyyyyyyyyxx.
So it definitively shows that the writes are not atomic. In my case 16byte writes were always atomic.
The answer was: POSIX doesn't mandate atomicity for writes/reads except for pipes. The 16 byte atomicity that I saw was kernel specific and may/can change in future.
Details of the answer in the actual post:
write(2)/read(2) atomicity between processes in linux
I am familiar with theory about filesystems in general, not with implementation of Ext4. Take this as educated guess.
Yes, I believe one sector reads and writes will be atomic because
Link you provided quotes "Currently concurrent reads/writes are atomic only wrt individual pages, however are not on the system call. "
Disk sector (512 bytes) writes are atomic according to Stephen Tweedie. In private email conversation with him, he acknowledged that this guarantee is only as good as the hardware.
Ext filesystems overwrite data in place, no copy on write. No allocation.
There is some effort to implement inline data, very small files data can fit in the inode itself. If you only need to store few bytes, that may have impact.
Not sure about one page, but it would make little sense in full journaling mode to send less than a page to the journal before commiting.
How guys from linux make /dev files. You can write to them and immediately they're erased.
I can imagine some program which constantly read some dev file:
FILE *fp;
char buffer[255];
int result;
fp = fopen(fileName, "r");
if (!fp) {
printf("Open file error");
return;
}
while (1)
{
result = fscanf(fp, "%254c", buffer);
printf("%s", buffer);
memset(buffer, 0, 255);
fflush(stdout);
sleep(1);
}
fclose(fp);
But how to delete content in there? Closing a file and opening them once again in "w" mode is not the way how they done it, because you can do i.e. cat > /dev/tty
What are files? Files are names in a directory structure which denote objects. When you open a file like /home/joe/foo.txt, the operating system creates an object in memory representing that file (or finds an existing one, if the file is already open), binds a descriptor to it which is returned and then operations on that file descriptor (like read and write) are directed, through the object, into file system code which manipulates the file's representation on disk.
Device entries are also names in the directory structure. When you open some /dev/foo, the operating system creates an in-memory object representing the device, or finds an existing one (in which case there may be an error if the device does not support multiple opens!). If successful, it binds a new file descriptor to the device obejct and returns that descriptor to your program. The object is configured in such a way that the operations like read and write on the descriptor are directed to call into the specific device driver for device foo, and correspond to doing some kind of I/O with that device.
Such entries in /dev/ are not files; a better name for them is "device nodes" (a justification for which is the name of the mknod command). Only when programmers and sysadmins are speaking very loosely do they call them "device files".
When you do cat > /dev/tty, there isn't anything which is "erasing" data "on the other end". Well, not exactly. Basically, cat is calling write on a descriptor, and this results in a chain of function calls which ends up somewhere in the kernel's tty subsystem. The data is handed off to a tty driver which will send the data into a serial port, or socket, or into a console device which paints characters on the screen or whatever. Virtual terminals like xterm use a pair of devices: a master and slave pseudo-tty. If a tty is connected to a pseudo-tty device, then cat > /dev/tty writes go through a kind of "trombone": they bubble up on the master side of the pseudo-tty, where in fact there is a while (1) loop in some user-space C program receiving the bytes, like from a pipe. That program is xterm (or whatever); it removes the data and draws the characters in its window, scrolls the window, etc.
Unix is designed so that devices (tty, printer, etc) are accessed like everything else (as a file) so the files in /dev are special pseudo files that represent the device within the file-system.
You don't want to delete the contents of such a device file, and honestly it could be dangerous for your system if you write to them willy-nilly without understanding exactly what you are doing.
Device files are not normal files, if "normal file" refers to an arbitrary sequence of bytes, often stored on a medium. But not all files are normal files.
More broadly, files are an abstraction referring to a system service and/or resource, a service being something you can send information to for some purpose (e.g., for a normal file, write data to storage) and a resource being something you request data from for some purpose (e.g., for a normal file, read data from storage). C defines a standard for interfacing with such a service/resource.
Device files fit within this definition, but they do not not necessarily match my more specific "normal file" examples of reading and writing to and from storage. You can directly create dev files, but the only meaningful reason to do so is within the context of a kernel module. More often you may refer to them (e.g., with udev), keeping in mind they are actually created by the kernel and represent an interface with the kernel. Beyond that, the functioning of the interface differs from dev file to dev file.
I've also found quiet nice explanation:
http://lwn.net/images/pdf/LDD3/ch18.pdf