mmap CMA area on /dev/mem - linux

I need reserve 256-512 Mb of continuous physical memory and have access to this memory from the user space.
I decided to use CMA for memory reserving.
Here are the steps on my idea that must be performed:
Reservation required amount of memory by CMA during system booting.
Parsing of CMA patch output which looks like for example: "CMA: reserved 256 MiB at 27400000" and saving two parameters: size of CMA area = 256*1024*1024 bytes and phys address of CMA area = 0x27400000.
Mapping of CMA area at /dev/mem file with offset = 0x27400000 using mmap(). (Of course, CONFIG_STRICT_DEVMEM is disabled)
It would let me to read data directly from phys memory from user space.
But the next code make segmentation fault(there size = 1Mb):
int file;
void* start;
file=open("/dev/mem", O_RDWR | O_SYNC);
if ( (start = mmap(0, 1024*1024, PROT_READ | PROT_WRITE, MAP_SHARED, file, 0x27400000)) == MAP_FAILED ){
perror("mmap");
}
for (int offs = 0; offs<50; offs++){
cout<<((char *)start)[offs];
}
Output of this code: "mmap: Invalid argument".
When I changed offset = 0x27400000 on 0, this code worked fine and program displayed trash. It also work for alot of offsets which I looked at /proc/iomem.
According to information from /proc/iomem, phys addr of CMA area (0x27400000 on my system) always situated in System RAM.
Does anyone have any ideas, how to mmap CMA area on /dev/mem? What am I doing wrong?
Thanks alot for any help!

Solution to this problem was suggested to me by Jeff Haran in kernelnewbies mailing list.
It was necessary to disable CONFIG_x86_PAT in .config and mmap() has started to work!
If CONFIG_X86_PAT is configured you will have problems mapping memory to user space. It basically implements the same restrictions as CONFIG_STRICT_DEVMEM.
Jeff Haran
Now I can mmap /dev/mem on any physical address I want.
But necessary to be careful:
Word of caution. CONFIG_X86_PAT was likely introduced for a reason. There may be some performance penalties for turning this off, though in my testing so far turning if off doesn’t seem to break anything.
Jeff Haran

Related

Why shouldn't I use ioremap on system memory for ARMv6+?

I need to reserve a large buffer of physically contiguous RAM from the kernel and be able to gaurantee that the buffer will always use a specific, hard-coded physical address. This buffer should remain reserved for the kernel's entire lifetime. I have written a chardev driver as an interface for accessing this buffer in userspace. My platform is an embedded system with ARMv7 architecture running a 2.6 Linux kernel.
Chapter 15 of Linux Device Drivers, Third Edition has the following to say on the topic (page 443):
Reserving the top of RAM is accomplished by passing a mem= argument to the kernel at boot time. For example, if you have 256 MB, the argument mem=255M keeps the kernel from using the top megabyte. Your module could later use the following code to gain access to such memory:
dmabuf = ioremap (0xFF00000 /* 255M */, 0x100000 /* 1M */);
I've done that plus a couple of other things:
I'm using the memmap bootarg in addition to the mem one. The kernel boot parameters documentation suggests always using memmap whenever you use mem to avoid address collisions.
I used request_mem_region before calling ioremap and, of course, I check that it succeeds before moving ahead.
This is what the system looks like after I've done all that:
# cat /proc/cmdline
root=/dev/mtdblock2 console=ttyS0,115200 init=/sbin/preinit earlyprintk debug mem=255M memmap=1M$255M
# cat /proc/iomem
08000000-0fffffff : PCIe Outbound Window, Port 0
08000000-082fffff : PCI Bus 0001:01
08000000-081fffff : 0001:01:00.0
08200000-08207fff : 0001:01:00.0
18000300-18000307 : serial
18000400-18000407 : serial
1800c000-1800cfff : dmu_regs
18012000-18012fff : pcie0
18013000-18013fff : pcie1
18014000-18014fff : pcie2
19000000-19000fff : cru_regs
1e000000-1fffffff : norflash
40000000-47ffffff : PCIe Outbound Window, Port 1
40000000-403fffff : PCI Bus 0002:01
40000000-403fffff : 0002:01:00.0
40400000-409fffff : PCI Bus 0002:01
40400000-407fffff : 0002:01:00.0
40800000-40807fff : 0002:01:00.0
80000000-8fefffff : System RAM
80052000-8045dfff : Kernel text
80478000-80500143 : Kernel data
8ff00000-8fffffff : foo
Everything so far looks good, and my driver is working perfectly. I'm able to read and write directly to the specific physical address I've chosen.
However, during bootup, a big scary warning (™) was triggered:
BUG: Your driver calls ioremap() on system memory. This leads
to architecturally unpredictable behaviour on ARMv6+, and ioremap()
will fail in the next kernel release. Please fix your driver.
------------[ cut here ]------------
WARNING: at arch/arm/mm/ioremap.c:211 __arm_ioremap_pfn_caller+0x8c/0x144()
Modules linked in:
[] (unwind_backtrace+0x0/0xf8) from [] (warn_slowpath_common+0x4c/0x64)
[] (warn_slowpath_common+0x4c/0x64) from [] (warn_slowpath_null+0x1c/0x24)
[] (warn_slowpath_null+0x1c/0x24) from [] (__arm_ioremap_pfn_caller+0x8c/0x144)
[] (__arm_ioremap_pfn_caller+0x8c/0x144) from [] (__arm_ioremap_caller+0x50/0x58)
[] (__arm_ioremap_caller+0x50/0x58) from [] (foo_init+0x204/0x2b0)
[] (foo_init+0x204/0x2b0) from [] (do_one_initcall+0x30/0x19c)
[] (do_one_initcall+0x30/0x19c) from [] (kernel_init+0x154/0x218)
[] (kernel_init+0x154/0x218) from [] (kernel_thread_exit+0x0/0x8)
---[ end trace 1a4cab5dbc05c3e7 ]---
Triggered from: arc/arm/mm/ioremap.c
/*
* Don't allow RAM to be mapped - this causes problems with ARMv6+
*/
if (pfn_valid(pfn)) {
printk(KERN_WARNING "BUG: Your driver calls ioremap() on system memory. This leads\n"
KERN_WARNING "to architecturally unpredictable behaviour on ARMv6+, and ioremap()\n"
KERN_WARNING "will fail in the next kernel release. Please fix your driver.\n");
WARN_ON(1);
}
What problems, exactly, could this cause? Can they be mitigated? What are my alternatives?
So I've done exactly that, and it's working.
Provide the kernel command line (e.g. /proc/cmdline) and the resulting memory map (i.e. /proc/iomem) to verify this.
What problems, exactly, could this cause?
The problem with using ioremap() on system memory is that you end up assigning conflicting attributes to the memory which causes "unpredictable" behavior.
See the article "ARM's multiply-mapped memory mess", which provides a history to the warning that you are triggering.
The ARM kernel maps RAM as normal memory with writeback caching; it's also marked non-shared on uniprocessor systems. The ioremap() system call, used to map I/O memory for CPU use, is different: that memory is mapped as device memory, uncached, and, maybe, shared. These different mappings give the expected behavior for both types of memory. Where things get tricky is when somebody calls ioremap() to create a new mapping for system RAM.
The problem with these multiple mappings is that they will have differing attributes. As of version 6 of the ARM architecture, the specified behavior in that situation is "unpredictable."
Note that "system memory" is the RAM that is managed by the kernel.
The fact that you trigger the warning indicates that your code is generating multiple mappings for a region of memory.
Can they be mitigated?
You have to ensure that the RAM you want to ioremap() is not "system memory", i.e. managed by the kernel.
See also this answer.
ADDENDUM
This warning that concerns you is the result of pfn_valid(pfn) returning TRUE rather than FALSE.
Based on the Linux cross-reference link that you provided for version 2.6.37,
pfn_valid() is simply returning the result of
memblock_is_memory(pfn << PAGE_SHIFT);
which in turn is simply returning the result of
memblock_search(&memblock.memory, addr) != -1;
I suggest that the kernel code be hacked so that the conflict is revealed.
Before the call to ioremap(), assign TRUE to the global variable memblock_debug.
The following patch should display the salient information about the memory conflict.
(The memblock list is ordered by base-address, so memblock_search() performs a binary search on this list, hence the use of mid as the index.)
static int __init_memblock memblock_search(struct memblock_type *type, phys_addr_t addr)
{
unsigned int left = 0, right = type->cnt;
do {
unsigned int mid = (right + left) / 2;
if (addr < type->regions[mid].base)
right = mid;
else if (addr >= (type->regions[mid].base +
type->regions[mid].size))
left = mid + 1;
- else
+ else {
+ if (memblock_debug)
+ pr_info("MATCH for 0x%x: m=0x%x b=0x%x s=0x%x\n",
+ addr, mid,
+ type->regions[mid].base,
+ type->regions[mid].size);
return mid;
+ }
} while (left < right);
return -1;
}
If you want to see all the memory blocks, then call memblock_dump_all() with the variable memblock_debug is TRUE.
[Interesting that this is essentially a programming question, yet we haven't seen any of your code.]
ADDENDUM 2
Since you're probably using ATAGs (instead of Device Tree), and you want to dedicate a memory region, fix up the ATAG_MEM to reflect this smaller size of physical memory.
Assuming you have made zero changes to your boot code, the ATAG_MEM is still specifying the full RAM, so perhaps this could be the source of the system memory conflict that causes the warning.
See this answer about ATAGs and this related answer.

Accessing memory pointers in hardware registers

I'm working on enhancing the stock ahci driver provided in Linux in order to perform some needed tasks. I'm at the point of attempting to issue commands to the AHCI HBA for the hard drive to process. However, whenever I do so, my system locks up and reboots. Trying to explain the process of issuing a command to an AHCI drive is far to much for this question. If needed, reference this link for the full discussion (the process is rather defined all over because there are several pieces, however, ch 4 has the data structures necessary).
Essentially, one writes the appropriate structures into memory regions defined by either the BIOS or the OS. The first memory region I should write to is the Command List Base Address contained in the register PxCLB (and PxCLBU if 64-bit addressing applies). My system is 64 bits and so I'm trying to getting both 32-bit registers. My code is essentially this:
void __iomem * pbase = ahci_port_base(ap);
u32 __iomem *temp = (u32*)(pbase + PORT_LST_ADDR);
struct ahci_cmd_hdr *cmd_hdr = NULL;
cmd_hdr = (struct ahci_cmd_hdr*)(u64)
((u64)(*(temp + PORT_LST_ADDR_HI)) << 32 | *temp);
pr_info("%s:%d cmd_list is %p\n", __func__, __LINE__, cmd_hdr);
// problems with this next line, makes the system reboot
//pr_info("%s:%d cl[0]:0x%08x\n", __func__, __LINE__, cmd_hdr->opts);
The function ahci_port_base() is found in the ahci driver (at least it is for CentOS 6.x). Basically, it returns the proper address for that port in the AHCI memory region. PORT_LST_ADDR and PORT_LST_ADDR_HI are both macros defined in that driver. The address that I get after getting both the high and low addresses is usually something like 0x0000000037900000. Is this memory address in a space that I cannot simply dereference it?
I'm hitting my head against the wall at this point because this link shows that accessing it in this manner is essentially how it's done.
The address that I get after getting both the high and low addresses
is usually something like 0x0000000037900000. Is this memory address
in a space that I cannot simply dereference it?
Yes, you are correct - that's a bus address, and you can't just dereference it because paging is enabled. (You shouldn't be just dereferencing the iomapped addresses either - you should be using readl() / writel() for those, but the breakage here is more subtle).
It looks like the right way to access the ahci_cmd_hdr in that driver is:
struct ahci_port_priv *pp = ap->private_data;
cmd_hdr = pp->cmd_slot;

Writing out DMA buffers into memory mapped file

I need to write in embedded Linux(2.6.37) as fast as possible incoming DMA buffers to HD partition as raw device /dev/sda1. Buffers are aligned as required and are of equal 512KB length. The process may continue for a very long time and fill as much as, for example, 256GB of data.
I need to use the memory-mapped file technique (O_DIRECT not applicable), but can't understand the exact way how to do this.
So, in pseudo code "normal" writing:
fd=open(/dev/sda1",O_WRONLY);
while(1) {
p = GetVirtualPointerToNewBuffer();
if (InputStopped())
break;
write(fd, p, BLOCK512KB);
}
Now, I will be very thankful for the similar pseudo/real code example of how to utilize memory-mapped technique for this writing.
UPDATE2:
Thanks to kestasx the latest working test code looks like following:
#define TSIZE (64*KB)
void* TBuf;
int main(int argc, char **argv) {
int fdi=open("input.dat", O_RDONLY);
//int fdo=open("/dev/sdb2", O_RDWR);
int fdo=open("output.dat", O_RDWR);
int i, offs=0;
void* addr;
i = posix_memalign(&TBuf, TSIZE, TSIZE);
if ((fdo < 1) || (fdi < 1)) {
printf("Error in files\n");
return -1; }
while(1) {
addr = mmap((void*)TBuf, TSIZE, PROT_READ | PROT_WRITE, MAP_SHARED, fdo, offs);
if ((unsigned int)addr == 0xFFFFFFFFUL) {
printf("Error MMAP=%d, %s\n", errno, strerror(errno));
return -1; }
i = read(fdi, TBuf, TSIZE);
if (i != TSIZE) {
printf("End of data\n");
return 0; }
i = munmap(addr, TSIZE);
offs += TSIZE;
sleep(1);
};
}
UPDATE3:
1. To precisely imitate the DMA work, I need to move read() call before mmp(), because when the DMA finishes it provides me with the address where it has put data. So, in pseudo code:
while(1) {
read(fdi, TBuf, TSIZE);
addr = mmap((void*)TBuf, TSIZE, PROT_READ|PROT_WRITE, MAP_FIXED|MAP_SHARED, fdo, offs);
munmap(addr, TSIZE);
offs += TSIZE; }
This variant fails after(!) the first loop - read() says BAD ADDRESS on TBuf.
Without understanding exactly what I do, I substituted munmap() with msync(). This worked perfectly.
So, the question here - why unmapping the addr influenced on TBuf?
2.With the previous example working I went to the real system with the DMA. The same loop, just instead of read() call is the call which waits for a DMA buffer to be ready and its virtual address provided.
There are no error, the code runs, BUT nothing is recorded (!).
My thought was that Linux does not see that the area was updated and therefore does not sync() a thing.
To test this, I eliminated in the working example the read() call - and yes, nothing was recorded too.
So, the question here - how can I tell Linux that the mapped region contains new data, please, flush it!
Thanks a lot!!!
If I correctly understand, it makes sense if You mmap() file (not sure if it You can mmap() raw partition/block-device) and data via DMA is written directly to this memory region.
For this to work You need to be able to control p (where new buffer is placed) or address where file is maped. If You don't - You'll have to copy memory contents (and will lose some benefits of mmap).
So psudo code would be:
truncate("data.bin", 256GB);
fd = open( "data.bin", O_RDWR );
p = GetVirtualPointerToNewBuffer();
adr = mmap( p, 1GB, PROT_READ | PROT_WRITE, MAP_SHARED, fd, offset_in_file );
startDMA();
waitDMAfinish();
munmap( adr, 1GB );
This is first step only and I'm not completely sure if it will work with DMA (have no such experience).
I assume it is 32bit system, but even then 1GB mapped file size may be too big (if Your RAM is smaller You'll be swaping).
If this setup will work, next step would be to make loop to map regions of file at different offsets and unmap already filled ones.
Most likely You'll need to align addr to 4KB boundary.
When You'll unmap region, it's data will be synced to disk. So You'll need some testing to select appropriate mapped region size (while next region is filled by DMA, there must be enough time to unmap/write previous one).
UPDATE:
What exactly happens when You fill mmap'ed region via DMA I simply don't know (not sure how exactly dirty pages are detected: what is done by hardware, and what must be done by software).
UPDATE2: To my best knowledge:
DMA works the following way:
CPU arranges DMA transfer (address where to write transfered data in RAM);
DMA controller does the actual work, while CPU can do it's own work in parallel;
once DMA transfer is complete - DMA controller signals CPU via IRQ line (interrupt), so CPU can handle the result.
This seems simple while virtual memory is not involved: DMA should work independently from runing process (actual VM table in use by CPU). Yet it should be some mehanism to invalidate CPU cache for modified by DMA physical RAM pages (don't know if CPU needs to do something, or it is done authomatically by hardware).
mmap() forks the following way:
after successfull call of mmap(), file on disk is attached to process memory range (most likely some data structure is filled in OS kernel to hold this info);
I/O (reading or writing) from mmaped range triggers pagefault, which is handled by kernel loading appropriate blocks from atached file;
writes to mmaped range are handled by hardware (don't know how exactly: maybe writes to previously unmodified pages triger some fault, which is handled by kernel marking these pages dirty; or maybe this marking is done entirely in hardware and this info is available to kernel when it needs to flush modified pages to disk).
modified (dirty) pages are written to disk by OS (as it sees appropriate) or can be forced via msync() or munmap()
In theory it should be possible to do DMA transfers to mmaped range, but You need to find out, how exactly pages ar marked dirty (if You need to do something to inform kernel which pages need to be written to disk).
UPDATE3:
Even if modified by DMA pages are not marked dirty, You should be able to triger marking by rewriting (reading ant then writing the same) at least one value in each page (most likely each 4KB) transfered. Just make sure this rewriting is not removed (optimised out) by compiler.
UPDATE4:
It seems file opened O_WRONLY can't be mmap'ed (see question comments, my experimets confirm this too). It is logical conclusion of mmap() workings described above. The same is confirmed here (with reference to POSIX standart requirement to ensure file is readable regardless of maping protection flags).
Unless there is some way around, it actually means that by using mmap() You can't avoid reading of results file (unnecessary step in Your case).
Regarding DMA transfers to mapped range, I think it will be a requirement to ensure maped pages are preloalocated before DMA starts (so there is real memory asigned to both DMA and maped region). On Linux there is MAP_POPULATE mmap flag, but from manual it seams it works with MAP_PRIVATE mapings only (changes are not writen to disk), so most likely it is usuitable. Likely You'll have to triger pagefaults manually by accessing each maped page. This should triger reading of results file.
If You still wish to use mmap and DMA together, but avoid reading of results file, You'll have to modify kernel internals to allow mmap to use O_WRONLY files (for example by zero-filling trigered pages, instead of reading them from disk).

Mmap is not working for high address memory mapping?

I am trying to do
memory = (char *)mmap((void *)0X0000100000000000,(size_t)0xffffffff/8,PROT_READ | PROT_WRITE , MAP_SHARED|MAP_ANONYMOUS,4,0);
but its not mapping anything and returning 0. I need to map memory at high address in 64-bit machine.
This is not meant as a complete answer - more of a possible explanation:
0X0000100000000000 is 281474976710656. Do you have that high a virtual memory address available? Or stated another way: is that address valid in your OS? I would guess the answer is no.
Is mmap actually returning MAP_FAILED ( (void *) -1 )? Usually when you give mmap an address it does not like, you get MAP_FAILED and errno == EINVAL. Did you check errno?
Note: 4 bytes is not the word length in a 64 bit OS, usually it is 8. A 4 byte word cannot address all of memory, for example.

After mmap(), write to returned address is OK, but read cause system crash.Why?

I want to share memory between two process.
After mmap(), I get a address mapStart, then I add offset to mapStart and get mapAddr, and make sure mapAddr will not exceed maped PAGE_SIZE.
When I write to mapAddr by
memcpy((void *)mapAddr, data, size);
everything is OK.
But when I read from mapAddr by
memcpy( &data, (void *)mapAddr, size);`
that will case system crash.
Who know Why?
The similar problem is here
Add some Info: #Tony Delroy, #J-16 SDiZ
mmap function is:
mapStart = (void volatile *)mmap(0, PAGE_SIZE, PROT_READ | PROT_WRITE, MAP_SHARED | MAP_LOCKED, memfd, pa_base);
system crash: have no any OS error message, Console print some MCA info
the detail described in here
Just some idea.
Is your mmap() spanning over memory regions with different attribute? This is illegal.
Older kernel (you said 2.6.18) allowed this, but crash when you write to some of it.
See this post for some starting point. If it is possible, try a newer kernel.
There are at least two possible issues:
After mmap(), I get a address mapStart, then I add offset to mapStart and get mapAddr, and make sure mapAddr will not exceed maped PAGE_SIZE.
Not mapAddr must be made sure not to exceed the mapped size, but mapAddr+size. You are trying to touch size bytes, not just one.
memcpy((void *)mapAddr, data, size);
memcpy( &data, (void *)mapAddr, size);
Assuming data is not a array (which is a plausible assumption since you use it without address operator in the first line), the second line copies not from the location pointed to by data, but starting with data. This is quite possibly some unallocated memory, or some location on the stack, or whatever. If there is not a lot on the stack, it might as well read beyond the end of the stack into the text segment, or... something else.
(If data is indeed an array, it is of course equivalent, but then your code style would be inconsistent.)

Resources