----- Original Message -----
>
> I see what you are concerning about...
> Attach the new format patch, which would just error out when it detect
> datatype failed.
>
> Thanks,
> Lei
I'm going to NAK this patch for three reasons.
First, there are literally hundreds of text symbols that do
not have any debuginfo data for one reason or another, i.e.,
they may be in .S assembly language files, or perhaps are
crafted with a macro, etc.
Without your patch, gdb at least informs you that they are
a text symbol:
crash> whatis sk_load_word
<text variable, no debug info> sk_load_word;
crash>
But your patch falsely proclaims that gdb failed to recognize it:
crash> whatis sk_load_word
whatis: gdb request failed: whatis sk_load_word
crash>
Secondly, there can be strange/invalid output on certain functions.
For example, without your patch:
crash> whatis alloc_iommu
int alloc_iommu(struct dmar_drhd_unit *);
crash>
But with your patch applied, it shows this:
crash> whatis alloc_iommu
int alloc_iommu(struct dmar_drhd_unit * dev size align_mask);
crash>
And that seems to be because there are two functions with
the same name:
crash> sym alloc_iommu
ffffffff81045950 (t) alloc_iommu /usr/src/debug/kernel-3.7.fc17/linux-3.7.9-104.fc17.x86_64/arch/x86/kernel/amd_gart_64.c: 96
ffffffff814fe920 (T) alloc_iommu /usr/src/debug/kernel-3.7.fc17/linux-3.7.9-104.fc17.x86_64/drivers/iommu/dmar.c: 644
crash>
And your command is mixing the prototype info of the two
functions:
static unsigned long alloc_iommu(struct device *dev, int size,
unsigned long align_mask)
int alloc_iommu(struct dmar_drhd_unit *drhd)
And third, there is still the issue with module text symbols
not working all of the time. I don't know why that happens,
but it's bothersome.
However, I will apply the simple changes that you requested for
gdb_get_datatype() so that you can keep a private branch with
your full patch in place.
Dave
--
Crash-utility mailing list
Crash-utility@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/crash-utility
