Re: [PATCH v2 2/3] platform/x86/intel/tpmi: Add debugfs interface
From: srinivas pandruvada
Date: Wed Jul 12 2023 - 19:08:59 EST
On Wed, 2023-07-12 at 18:13 +0300, Andy Shevchenko wrote:
> On Tue, Jul 11, 2023 at 03:09:48PM -0700, Srinivas Pandruvada wrote:
> > Add debugfs interface for debugging TPMI configuration and register
> > contents. This shows PFS (PM Feature structure) for each TPMI
> > device.
> >
> > For each feature, show full register contents and allow to modify
> > register at an offset.
> >
> > This debugfs interface is not present on locked down kernel with no
> > DEVMEM access and without CAP_SYS_RAWIO permission.
>
> ...
>
> > struct intel_tpmi_pm_feature {
> > struct intel_tpmi_pfs_entry pfs_header;
> > unsigned int vsec_offset;
> > + struct intel_vsec_device *vsec_dev;
>
> Hmm... I don't know the layout of pfs_header, but this may be 4 bytes
> less
> if you move it upper.
The pfs_header is packed with size of 64 bit. So size will not change.
>
> > };
>
> ...
>
> > + for (count = 0; count < pfs->pfs_header.num_entries;
> > ++count) {
>
> > + size = pfs->pfs_header.entry_size * sizeof(u32);
>
> You already used this once, perhaps a macro helper?
> Also you can add there a comment that this comes from the trusted hw.
>
Added.
> > + /* The size is from a trusted hardware, but verify
> > anyway */
> > + if (size > TPMI_MAX_BUFFER_SIZE) {
> > + /*
> > + * The next offset depends on the current
> > size. So, can't skip to the
> > + * display of the next entry. Simply return
> > from this function with error.
> > + */
> > + ret = -EIO;
> > + goto done_mem_show;
> > + }
> > +
> > + buffer = kmalloc(size, GFP_KERNEL);
> > + if (!buffer) {
> > + ret = -ENOMEM;
> > + goto done_mem_show;
> > + }
> > +
> > + seq_printf(s, "TPMI Instance:%d offset:0x%x\n",
> > count, off);
> > +
> > + mem = ioremap(off, size);
> > + if (!mem) {
> > + ret = -ENOMEM;
> > + kfree(buffer);
> > + goto done_mem_show;
> > + }
> > +
> > + memcpy_fromio(buffer, mem, size);
> > +
> > + seq_hex_dump(s, " ", DUMP_PREFIX_OFFSET, row_size,
> > sizeof(u32), buffer, size,
> > + false);
> > +
> > + iounmap(mem);
> > + kfree(buffer);
> > +
> > + off += size;
> > + }
> > +
> > +done_mem_show:
> > + mutex_unlock(&tpmi_dev_lock);
> > +
> > + return ret;
> > +}
>
> ...
>
> > + size = pfs->pfs_header.entry_size * sizeof(u32);
> > + if (size > TPMI_MAX_BUFFER_SIZE)
> > + return -EIO;
>
> Again a dup even with a check.
>
> ...
>
> > + top_dir = debugfs_create_dir(name, NULL);
> > + if (IS_ERR_OR_NULL(top_dir))
>
> I dunno if I told you, but after a discussion (elsewhere), I realized
> two things:
> 1) this one never returns NULL;
> 2) even if error pointer is returned, the debugfs API is aware and
> will do nothing.
>
> Hence this conditional is redundant.
Removed that. My original version didn't check the return value.
>
> > + return;
>
> ...
>
> > + for (i = 0; i < tpmi_info->feature_count; ++i) {
>
> Why preincrement?
Does it matter for a "for" loop increment?
Thanks,
Srinivas
>
> > + struct intel_tpmi_pm_feature *pfs;
> > + struct dentry *dir;
> > +
> > + pfs = &tpmi_info->tpmi_features[i];
> > + snprintf(name, sizeof(name), "tpmi-id-%02x", pfs-
> > >pfs_header.tpmi_id);
> > + dir = debugfs_create_dir(name, top_dir);
> > +
> > + debugfs_create_file("mem_dump", 0444, dir, pfs,
> > &tpmi_mem_dump_fops);
> > + debugfs_create_file("mem_write", 0644, dir, pfs,
> > &mem_write_ops);
> > + }
>