Re: [Qemu-devel] [PATCH 1/7] vnvram: VNVRAM bdrv support

[Corey Bryant <coreyb@linux.vnet.ibm.com>]

On 05/24/2013 09:06 AM, Kevin Wolf wrote:
> Am 23.05.2013 um 19:44 hat Corey Bryant geschrieben:
>> Provides low-level VNVRAM functionality that reads and writes data,
>> such as an entry's binary blob, to a drive image using the block
>> driver.
>>
>> Signed-off-by: Corey Bryant <coreyb@linux.vnet.ibm.com>
>
>> +/*
>> + * Increase the drive size if it's too small to fit the VNVRAM data
>> + */
>> +static int vnvram_drv_adjust_size(VNVRAM *vnvram)
>> +{
>> +    int rc = 0;
>> +    int64_t needed_size;
>> +
>> +    needed_size = 0;
>> +
>> +    if (bdrv_getlength(vnvram->bds) < needed_size) {
>> +        rc = bdrv_truncate(vnvram->bds, needed_size);
>> +        if (rc != 0) {
>> +            DPRINTF("%s: VNVRAM drive too small\n", __func__);
>> +        }
>> +    }
>> +
>> +    return rc;
>> +}
>
> This function doesn't make a whole lot of sense. It truncates the file
> to size 0 if and only if bdrv_getlength() returns an error.
>

There's a later patch that adds a "get size" function and changes the 
initialization of needed_size to the actual size needed to store VNVRAM 
data.  Anyway I should probably just include that change in this patch. 
  I think I'll still need this function or part of it with the new 
simplified approach that it looks like we're going to take.

>> +
>> +/*
>> + * Write a header to the drive with entry count of zero
>> + */
>> +static int vnvram_drv_hdr_create_empty(VNVRAM *vnvram)
>> +{
>> +    VNVRAMDrvHdr hdr;
>> +
>> +    hdr.version = VNVRAM_CURRENT_VERSION;
>> +    hdr.magic = VNVRAM_MAGIC;
>> +    hdr.num_entries = 0;
>> +
>> +    vnvram_drv_hdr_cpu_to_be((&hdr));
>> +
>> +    if (bdrv_pwrite(vnvram->bds, 0, (&hdr), sizeof(hdr)) != sizeof(hdr)) {
>> +        DPRINTF("%s: Write of header to drive failed\n", __func__);
>> +        return -EIO;
>> +    }
>> +
>> +    vnvram->end_offset = sizeof(VNVRAMDrvHdr);
>> +
>> +    return 0;
>> +}
>> +
>> +/*
>> + * Read the header from the drive
>> + */
>> +static int vnvram_drv_hdr_read(VNVRAM *vnvram, VNVRAMDrvHdr *hdr)
>> +{
>> +    if (bdrv_pread(vnvram->bds, 0, hdr, sizeof(*hdr)) != sizeof(*hdr)) {
>> +        DPRINTF("%s: Read of header from drive failed\n", __func__);
>> +        return -EIO;
>> +    }
>
> Why do you turn all errors into -EIO instead of returning the real error
> code? (More instances of the same thing follow)
>

Good point, there's no reason to mask the original error code.

>> +
>> +    vnvram_drv_hdr_be_to_cpu(hdr);
>> +
>> +    return 0;
>> +}
>> +}
>
> Kevin
>
>
>

-- 
Regards,
Corey Bryant