Skip to contentSkip to conversation

parisc: Remove memcpy_fromio

PATCH In review Julian Vetter 3 participants 5 messages Jan 30, 2025 – Feb 3, 2025

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 parisc: Remove memcpy_fromio

Fully migrate parisc to the IO functions from lib/iomem_copy.c. In a recent patch the functions memset_io and memcpy_toio were removed, but the memcpy_fromio was kept, because for very short sequences it does half word accesses, whereas the functions in lib/iomem_copy.c do byte accesses until the memory is naturally aligned and then do machine word accesses. But I don't think the single half-word access merits keeping the arch specific implementation, so, remove it as well.

---
 arch/parisc/include/asm/io.h      |  3 --
 arch/parisc/kernel/parisc_ksyms.c |  1 -
 arch/parisc/lib/io.c              | 61 -------------------------------
 3 files changed, 65 deletions(-)
arch/parisc/include/asm/io.h +0 -3
index 3143cf29ce27..61173a2b38e4 100644
@@ -135,9 +135,6 @@ static inline void gsc_writeq(unsigned long long val, unsigned long addr)
 
 #define pci_iounmap			pci_iounmap
 
-void memcpy_fromio(void *dst, const volatile void __iomem *src, int count);
-#define memcpy_fromio memcpy_fromio
-
 /* Port-space IO */
 
 #define inb_p inb
arch/parisc/kernel/parisc_ksyms.c +0 -1
index 1c366b0d3134..509146a52725 100644
@@ -43,7 +43,6 @@ EXPORT_SYMBOL($global$);
 #endif
 
 #include <asm/io.h>
-EXPORT_SYMBOL(memcpy_fromio);
 
 extern void $$divI(void);
 extern void $$divU(void);
arch/parisc/lib/io.c +0 -61

67 lines not on this page · raw message

Signature · 1 line
2.34.1
maintainer Acked-by on v1 1/1
#1 lore raw
Quoted 9 lines · Julian Vetter

Fully migrate parisc to the IO functions from lib/iomem_copy.c. In a recent patch the functions memset_io and memcpy_toio were removed, but the memcpy_fromio was kept, because for very short sequences it does half word accesses, whereas the functions in lib/iomem_copy.c do byte accesses until the memory is naturally aligned and then do machine word accesses. But I don't think the single half-word access merits keeping the arch specific implementation, so, remove it as well.

This one looks fairly obvious. It might be nice to also clean up the {in,out}s{b,w,l} helper functions in the same file, but I don't understand why those are special in the first place.

Those functions have been unchanged since before the git history and there are some comments that I don't find helpful. One thing they do is to deal with unaligned memory buffers, which the generic ones don't, but that could be easily added using get_unaligned/put_unaligned, expecting the compiler to optimize the memory side of the transfer.

Arnd

#2 lore raw
Quoted 9 lines · Arnd Bergmann
Quoted 8 lines · Julian Vetter

Fully migrate parisc to the IO functions from lib/iomem_copy.c. In a recent patch the functions memset_io and memcpy_toio were removed, but the memcpy_fromio was kept, because for very short sequences it does half word accesses, whereas the functions in lib/iomem_copy.c do byte accesses until the memory is naturally aligned and then do machine word accesses. But I don't think the single half-word access merits keeping the arch specific implementation, so, remove it as well.

Quoted 9 lines
Quoted 1 line

This one looks fairly obvious. It might be nice to also clean up the {in,out}s{b,w,l} helper functions in the same file, but I don't understand why those are special in the first place.

Not sure, either. But some comments contain: "just using the inlined version of the inw() breaks things". This is clear, and what you point out below already, these parisc specific functions deal with unaligned buffers, so using, e.g., the generic "inw()" will surely break things.

Quoted 6 lines

Those functions have been unchanged since before the git history and there are some comments that I don't find helpful. One thing they do is to deal with unaligned memory buffers, which the generic ones don't, but that could be easily added using get_unaligned/put_unaligned, expecting the compiler to optimize the memory side of the transfer.

I'm not sure what you suggest. Do you mean adding get_unaligned/put_unaligned in the generic include/asm-generic/io.h functions, or just adding get_unaligned/put_unaligned to the body of the parisc specific functions? Instead of all the "switch... case 0x2..." code?

Julian

Quoted 2 lines

Arnd

maintainer
#3 lore raw
Quoted 8 lines · Julian Vetter
Quoted 6 lines · Arnd Bergmann

Those functions have been unchanged since before the git history and there are some comments that I don't find helpful. One thing they do is to deal with unaligned memory buffers, which the generic ones don't, but that could be easily added using get_unaligned/put_unaligned, expecting the compiler to optimize the memory side of the transfer.

Quoted 4 lines

I'm not sure what you suggest. Do you mean adding get_unaligned/put_unaligned in the generic include/asm-generic/io.h functions, or just adding get_unaligned/put_unaligned to the body of the parisc specific functions? Instead of all the "switch... case 0x2..." code?

I meant adding put_unaligned/get_unaligned like this:

include/asm-generic/io.h +2 -1
@@ -422,7 +422,8 @@ static inline void readsw(const volatile void __iomem *addr, void *buffer,
 
                do {
                        u16 x = __raw_readw(addr);
-                       *buf++ = x;
+                       put_unaligned(x, buf);
+                       buf++;
                } while (--count);
        }
 }

This has no effect on architectures that can handle unaligned load/store on memory, and should behave the same way as the parisc code otherwise. On powerpc and possibly others, there needs to be a barrier (eieio() on powerpc) inside of the loop, according to Segher's earlier comments.

Arnd

maintainer
#4 lore raw
In reply to Arnd Bergmann
Quoted 9 lines · Arnd Bergmann
Quoted 8 lines · Julian Vetter

Fully migrate parisc to the IO functions from lib/iomem_copy.c. In a recent patch the functions memset_io and memcpy_toio were removed, but the memcpy_fromio was kept, because for very short sequences it does half word accesses, whereas the functions in lib/iomem_copy.c do byte accesses until the memory is naturally aligned and then do machine word accesses. But I don't think the single half-word access merits keeping the arch specific implementation, so, remove it as well.

Quoted 3 lines
Quoted 1 line

Thanks!
I've added the patch to the parisc git tree.

Helge