lib/zlib: add s390 hardware support for kernel zlib_deflate

Patch series "S390 hardware support for kernel zlib", v3.

With IBM z15 mainframe the new DFLTCC instruction is available.  It
implements deflate algorithm in hardware (Nest Acceleration Unit - NXU)
with estimated compression and decompression performance orders of
magnitude faster than the current zlib.

This patchset adds s390 hardware compression support to kernel zlib.
The code is based on the userspace zlib implementation:

	https://github.com/madler/zlib/pull/410

The coding style is also preserved for future maintainability.  There is
only limited set of userspace zlib functions represented in kernel.
Apart from that, all the memory allocation should be performed in
advance.  Thus, the workarea structures are extended with the parameter
lists required for the DEFLATE CONVENTION CALL instruction.

Since kernel zlib itself does not support gzip headers, only Adler-32
checksum is processed (also can be produced by DFLTCC facility).  Like
it was implemented for userspace, kernel zlib will compress in hardware
on level 1, and in software on all other levels.  Decompression will
always happen in hardware (when enabled).

Two DFLTCC compression calls produce the same results only when they
both are made on machines of the same generation, and when the
respective buffers have the same offset relative to the start of the
page.  Therefore care should be taken when using hardware compression
when reproducible results are desired.  However it does always produce
the standard conform output which can be inflated anyway.

The new kernel command line parameter 'dfltcc' is introduced to
configure s390 zlib hardware support:

    Format: { on | off | def_only | inf_only | always }
     on:       s390 zlib hardware support for compression on
               level 1 and decompression (default)
     off:      No s390 zlib hardware support
     def_only: s390 zlib hardware support for deflate
               only (compression on level 1)
     inf_only: s390 zlib hardware support for inflate
               only (decompression)
     always:   Same as 'on' but ignores the selected compression
               level always using hardware support (used for debugging)

The main purpose of the integration of the NXU support into the kernel
zlib is the use of hardware deflate in btrfs filesystem with on-the-fly
compression enabled.  Apart from that, hardware support can also be used
during boot for decompressing the kernel or the ramdisk image

With the patch for btrfs expanding zlib buffer from 1 to 4 pages (patch
6) the following performance results have been achieved using the
ramdisk with btrfs.  These are relative numbers based on throughput rate
and compression ratio for zlib level 1:

  Input data              Deflate rate   Inflate rate   Compression ratio
                          NXU/Software   NXU/Software   NXU/Software
  stream of zeroes        1.46           1.02           1.00
  random ASCII data       10.44          3.00           0.96
  ASCII text (dickens)    6,21           3.33           0.94
  binary data (vmlinux)   8,37           3.90           1.02

This means that s390 hardware deflate can provide up to 10 times faster
compression (on level 1) and up to 4 times faster decompression (refers
to all compression levels) for btrfs zlib.

Disclaimer: Performance results are based on IBM internal tests using DD
command-line utility on btrfs on a Fedora 30 based internal driver in
native LPAR on a z15 system.  Results may vary based on individual
workload, configuration and software levels.

This patch (of 9):

Create zlib_dfltcc library with the s390 DEFLATE CONVERSION CALL
implementation and related compression functions.  Update zlib_deflate
functions with the hooks for s390 hardware support and adjust workspace
structures with extra parameter lists required for hardware deflate.

Link: http://lkml.kernel.org/r/20200103223334.20669-2-zaslonko@linux.ibm.com
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Signed-off-by: Mikhail Zaslonko <zaslonko@linux.ibm.com>
Co-developed-by: Ilya Leoshkevich <iii@linux.ibm.com>
Cc: Chris Mason <clm@fb.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: David Sterba <dsterba@suse.com>
Cc: Eduard Shishkin <edward6@linux.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Josef Bacik <josef@toxicpanda.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

1 files changed, 273 insertions, 0 deletions

diff --git a/lib/zlib_dfltcc/dfltcc_deflate.c b/lib/zlib_dfltcc/dfltcc_deflate.c
new file mode 100644
index 000000000000..9f0f1c90c4e0
--- /dev/null
+++ b/lib/zlib_dfltcc/dfltcc_deflate.c
@@ -0,0 +1,273 @@
+// SPDX-License-Identifier: Zlib
+
+#include "../zlib_deflate/defutil.h"
+#include "dfltcc_util.h"
+#include "dfltcc.h"
+#include <linux/zutil.h>
+
+/*
+ * Compress.
+ */
+int dfltcc_can_deflate(
+    z_streamp strm
+)
+{
+    deflate_state *state = (deflate_state *)strm->state;
+    struct dfltcc_state *dfltcc_state = GET_DFLTCC_STATE(state);
+
+    /* Unsupported compression settings */
+    if (!dfltcc_are_params_ok(state->level, state->w_bits, state->strategy,
+                              dfltcc_state->level_mask))
+        return 0;
+
+    /* Unsupported hardware */
+    if (!is_bit_set(dfltcc_state->af.fns, DFLTCC_GDHT) ||
+            !is_bit_set(dfltcc_state->af.fns, DFLTCC_CMPR) ||
+            !is_bit_set(dfltcc_state->af.fmts, DFLTCC_FMT0))
+        return 0;
+
+    return 1;
+}
+
+static void dfltcc_gdht(
+    z_streamp strm
+)
+{
+    deflate_state *state = (deflate_state *)strm->state;
+    struct dfltcc_param_v0 *param = &GET_DFLTCC_STATE(state)->param;
+    size_t avail_in = avail_in = strm->avail_in;
+
+    dfltcc(DFLTCC_GDHT,
+           param, NULL, NULL,
+           &strm->next_in, &avail_in, NULL);
+}
+
+static dfltcc_cc dfltcc_cmpr(
+    z_streamp strm
+)
+{
+    deflate_state *state = (deflate_state *)strm->state;
+    struct dfltcc_param_v0 *param = &GET_DFLTCC_STATE(state)->param;
+    size_t avail_in = strm->avail_in;
+    size_t avail_out = strm->avail_out;
+    dfltcc_cc cc;
+
+    cc = dfltcc(DFLTCC_CMPR | HBT_CIRCULAR,
+                param, &strm->next_out, &avail_out,
+                &strm->next_in, &avail_in, state->window);
+    strm->total_in += (strm->avail_in - avail_in);
+    strm->total_out += (strm->avail_out - avail_out);
+    strm->avail_in = avail_in;
+    strm->avail_out = avail_out;
+    return cc;
+}
+
+static void send_eobs(
+    z_streamp strm,
+    const struct dfltcc_param_v0 *param
+)
+{
+    deflate_state *state = (deflate_state *)strm->state;
+
+    zlib_tr_send_bits(
+          state,
+          bi_reverse(param->eobs >> (15 - param->eobl), param->eobl),
+          param->eobl);
+    flush_pending(strm);
+    if (state->pending != 0) {
+        /* The remaining data is located in pending_out[0:pending]. If someone
+         * calls put_byte() - this might happen in deflate() - the byte will be
+         * placed into pending_buf[pending], which is incorrect. Move the
+         * remaining data to the beginning of pending_buf so that put_byte() is
+         * usable again.
+         */
+        memmove(state->pending_buf, state->pending_out, state->pending);
+        state->pending_out = state->pending_buf;
+    }
+#ifdef ZLIB_DEBUG
+    state->compressed_len += param->eobl;
+#endif
+}
+
+int dfltcc_deflate(
+    z_streamp strm,
+    int flush,
+    block_state *result
+)
+{
+    deflate_state *state = (deflate_state *)strm->state;
+    struct dfltcc_state *dfltcc_state = GET_DFLTCC_STATE(state);
+    struct dfltcc_param_v0 *param = &dfltcc_state->param;
+    uInt masked_avail_in;
+    dfltcc_cc cc;
+    int need_empty_block;
+    int soft_bcc;
+    int no_flush;
+
+    if (!dfltcc_can_deflate(strm))
+        return 0;
+
+again:
+    masked_avail_in = 0;
+    soft_bcc = 0;
+    no_flush = flush == Z_NO_FLUSH;
+
+    /* Trailing empty block. Switch to software, except when Continuation Flag
+     * is set, which means that DFLTCC has buffered some output in the
+     * parameter block and needs to be called again in order to flush it.
+     */
+    if (flush == Z_FINISH && strm->avail_in == 0 && !param->cf) {
+        if (param->bcf) {
+            /* A block is still open, and the hardware does not support closing
+             * blocks without adding data. Thus, close it manually.
+             */
+            send_eobs(strm, param);
+            param->bcf = 0;
+        }
+        return 0;
+    }
+
+    if (strm->avail_in == 0 && !param->cf) {
+        *result = need_more;
+        return 1;
+    }
+
+    /* There is an open non-BFINAL block, we are not going to close it just
+     * yet, we have compressed more than DFLTCC_BLOCK_SIZE bytes and we see
+     * more than DFLTCC_DHT_MIN_SAMPLE_SIZE bytes. Open a new block with a new
+     * DHT in order to adapt to a possibly changed input data distribution.
+     */
+    if (param->bcf && no_flush &&
+            strm->total_in > dfltcc_state->block_threshold &&
+            strm->avail_in >= dfltcc_state->dht_threshold) {
+        if (param->cf) {
+            /* We need to flush the DFLTCC buffer before writing the
+             * End-of-block Symbol. Mask the input data and proceed as usual.
+             */
+            masked_avail_in += strm->avail_in;
+            strm->avail_in = 0;
+            no_flush = 0;
+        } else {
+            /* DFLTCC buffer is empty, so we can manually write the
+             * End-of-block Symbol right away.
+             */
+            send_eobs(strm, param);
+            param->bcf = 0;
+            dfltcc_state->block_threshold =
+                strm->total_in + dfltcc_state->block_size;
+            if (strm->avail_out == 0) {
+                *result = need_more;
+                return 1;
+            }
+        }
+    }
+
+    /* The caller gave us too much data. Pass only one block worth of
+     * uncompressed data to DFLTCC and mask the rest, so that on the next
+     * iteration we start a new block.
+     */
+    if (no_flush && strm->avail_in > dfltcc_state->block_size) {
+        masked_avail_in += (strm->avail_in - dfltcc_state->block_size);
+        strm->avail_in = dfltcc_state->block_size;
+    }
+
+    /* When we have an open non-BFINAL deflate block and caller indicates that
+     * the stream is ending, we need to close an open deflate block and open a
+     * BFINAL one.
+     */
+    need_empty_block = flush == Z_FINISH && param->bcf && !param->bhf;
+
+    /* Translate stream to parameter block */
+    param->cvt = CVT_ADLER32;
+    if (!no_flush)
+        /* We need to close a block. Always do this in software - when there is
+         * no input data, the hardware will not nohor BCC. */
+        soft_bcc = 1;
+    if (flush == Z_FINISH && !param->bcf)
+        /* We are about to open a BFINAL block, set Block Header Final bit
+         * until the stream ends.
+         */
+        param->bhf = 1;
+    /* DFLTCC-CMPR will write to next_out, so make sure that buffers with
+     * higher precedence are empty.
+     */
+    Assert(state->pending == 0, "There must be no pending bytes");
+    Assert(state->bi_valid < 8, "There must be less than 8 pending bits");
+    param->sbb = (unsigned int)state->bi_valid;
+    if (param->sbb > 0)
+        *strm->next_out = (Byte)state->bi_buf;
+    if (param->hl)
+        param->nt = 0; /* Honor history */
+    param->cv = strm->adler;
+
+    /* When opening a block, choose a Huffman-Table Type */
+    if (!param->bcf) {
+        if (strm->total_in == 0 && dfltcc_state->block_threshold > 0) {
+            param->htt = HTT_FIXED;
+        }
+        else {
+            param->htt = HTT_DYNAMIC;
+            dfltcc_gdht(strm);
+        }
+    }
+
+    /* Deflate */
+    do {
+        cc = dfltcc_cmpr(strm);
+        if (strm->avail_in < 4096 && masked_avail_in > 0)
+            /* We are about to call DFLTCC with a small input buffer, which is
+             * inefficient. Since there is masked data, there will be at least
+             * one more DFLTCC call, so skip the current one and make the next
+             * one handle more data.
+             */
+            break;
+    } while (cc == DFLTCC_CC_AGAIN);
+
+    /* Translate parameter block to stream */
+    strm->msg = oesc_msg(dfltcc_state->msg, param->oesc);
+    state->bi_valid = param->sbb;
+    if (state->bi_valid == 0)
+        state->bi_buf = 0; /* Avoid accessing next_out */
+    else
+        state->bi_buf = *strm->next_out & ((1 << state->bi_valid) - 1);
+    strm->adler = param->cv;
+
+    /* Unmask the input data */
+    strm->avail_in += masked_avail_in;
+    masked_avail_in = 0;
+
+    /* If we encounter an error, it means there is a bug in DFLTCC call */
+    Assert(cc != DFLTCC_CC_OP2_CORRUPT || param->oesc == 0, "BUG");
+
+    /* Update Block-Continuation Flag. It will be used to check whether to call
+     * GDHT the next time.
+     */
+    if (cc == DFLTCC_CC_OK) {
+        if (soft_bcc) {
+            send_eobs(strm, param);
+            param->bcf = 0;
+            dfltcc_state->block_threshold =
+                strm->total_in + dfltcc_state->block_size;
+        } else
+            param->bcf = 1;
+        if (flush == Z_FINISH) {
+            if (need_empty_block)
+                /* Make the current deflate() call also close the stream */
+                return 0;
+            else {
+                bi_windup(state);
+                *result = finish_done;
+            }
+        } else {
+            if (flush == Z_FULL_FLUSH)
+                param->hl = 0; /* Clear history */
+            *result = flush == Z_NO_FLUSH ? need_more : block_done;
+        }
+    } else {
+        param->bcf = 1;
+        *result = need_more;
+    }
+    if (strm->avail_in != 0 && strm->avail_out != 0)
+        goto again; /* deflate() must use all input or all output */
+    return 1;
+}