qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [RFC PATCH] hw/arm/virt: use variable size of flash dev


From: Xiang Zheng
Subject: Re: [Qemu-devel] [RFC PATCH] hw/arm/virt: use variable size of flash device to save memory
Date: Mon, 8 Apr 2019 21:43:46 +0800
User-agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:64.0) Gecko/20100101 Thunderbird/64.0

On 2019/4/3 23:35, Laszlo Ersek wrote:
>> I thought about your comments and wrote the following patch (just for test)
>> which uses a file mapping to replace the anonymous mapping. UEFI seems to 
>> work
>> fine. So why not use a file mapping to read or write a pflash device?
> Honestly I can't answer "why not do this". Maybe we should.
> 
> Regarding "why not do this *exactly as shown below*" -- probably because
> then we'd have updates to the same underlying regular file via both
> mmap() and write(), and the interactions between those are really tricky
> (= best avoided).
> 
> One of my favorite questions over the years used to be posting a matrix
> of possible mmap() and file descriptor r/w/sync interactions, with the
> outcomes as they were "implied" by POSIX, and then asking at the bottom,
> "is my understanding correct?" I've never ever received an answer to
> that. :)

In my opinion, it's feasible to r/w/sync the memory devices which use a block
backend via mmap() and write().

> 
> Also... although we don't really use them in practice, some QCOW2
> features for pflash block backends are -- or would be -- nice. (Not the
> internal snapshots or the live RAM dumps, of course.)

Regarding sparse files, can we read actual backend size data for the read-only
pflash memory as the following steps shown?

1) Create a sparse file -- QEMU_EFI-test.raw:
   dd if=/dev/zero of=QEMU_EFI-test.raw bs=1M seek=64 count=0

2) Read from QEMU_EFI.fd and write to QEMU_EFI-test.raw:
   dd of="QEMU_EFI-test.raw" if="QEMU_EFI.fd" conv=notrunc

3) Use QEMU_EFI-test.raw as the read-only pflash and start qemu with the below
patch applied:

---8>---

diff --git a/hw/block/block.c b/hw/block/block.c
index bf56c76..ad18d5e 100644
--- a/hw/block/block.c
+++ b/hw/block/block.c
@@ -30,7 +30,7 @@
 bool blk_check_size_and_read_all(BlockBackend *blk, void *buf, hwaddr size,
                                  Error **errp)
 {
-    int64_t blk_len;
+    int64_t blk_len, actual_len;
     int ret;

     blk_len = blk_getlength(blk);
@@ -53,7 +53,9 @@ bool blk_check_size_and_read_all(BlockBackend *blk, void 
*buf, hwaddr size,
      * block device and read only on demand.
      */
     assert(size <= BDRV_REQUEST_MAX_BYTES);
-    ret = blk_pread(blk, 0, buf, size);
+    actual_len = bdrv_get_allocated_file_size(blk_bs(blk));
+    ret = blk_pread(blk, 0, buf,
+            (!blk_is_read_only(blk) || actual_len < 0) ? size : actual_len);
     if (ret < 0) {
         error_setg_errno(errp, -ret, "can't read block backend");
         return false;
-- 
1.8.3.1


-- 

Thanks,
Xiang





reply via email to

[Prev in Thread] Current Thread [Next in Thread]